No public description

PiperOrigin-RevId: 595039397

mediapipe/calculators/image/BUILD

@@ -301,9 +301,11 @@ cc_test(
         "//mediapipe/framework/port:parse_text_proto",
         "//mediapipe/gpu:gpu_buffer_to_image_frame_calculator",
         "//mediapipe/gpu:image_frame_to_gpu_buffer_calculator",
+        "//mediapipe/gpu:multi_pool",
         "//third_party:opencv",
         "@com_google_absl//absl/container:flat_hash_set",
         "@com_google_absl//absl/flags:flag",
+        "@com_google_absl//absl/log:absl_check",
         "@com_google_absl//absl/strings",
         "@com_google_googletest//:gtest_main",
     ],

mediapipe/calculators/image/image_transformation_calculator.cc

@@ -656,6 +656,15 @@ absl::Status ImageTransformationCalculator::RenderGpu(CalculatorContext* cc) {
                                                   input.format());
 
   gpu_helper_.BindFramebuffer(dst);
+
+  if (scale_mode_ == mediapipe::ScaleMode::FIT) {
+    // In kFit scale mode, the rendered quad does not fill the whole
+    // framebuffer, so clear it beforehand.
+    glClearColor(padding_color_[0] / 255.0f, padding_color_[1] / 255.0f,
+                 padding_color_[2] / 255.0f, 1.0f);
+    glClear(GL_COLOR_BUFFER_BIT);
+  }
+
   glActiveTexture(GL_TEXTURE1);
   glBindTexture(src1.target(), src1.name());
 

mediapipe/calculators/image/image_transformation_calculator.proto

@@ -46,13 +46,14 @@ message ImageTransformationCalculatorOptions {
   optional bool flip_horizontally = 5 [default = false];
   // Scale mode.
   optional ScaleMode.Mode scale_mode = 6;
-  // Padding type. This option is only used when the scale mode is FIT.
-  // Default is to use BORDER_CONSTANT. If set to false, it will use
-  // BORDER_REPLICATE instead.
+  // Padding type. This option is only used when the scale mode is FIT. If set
+  // to true (default), a constant border is added with color specified by
+  // padding_color. If set to false, a border is added by replicating edge
+  // pixels (only supported for CPU).
   optional bool constant_padding = 7 [default = true];
 
   // The color for the padding. This option is only used when the scale mode is
-  // FIT. Default is black. This is for CPU only.
+  // FIT. Default is black.
   optional Color padding_color = 8;
 
   // Interpolation method to use. Note that on CPU when LINEAR is specified,

mediapipe/calculators/image/image_transformation_calculator_test.cc

@@ -1,9 +1,11 @@
+#include <algorithm>
 #include <string>
 #include <utility>
 #include <vector>
 
 #include "absl/container/flat_hash_set.h"
 #include "absl/flags/flag.h"
+#include "absl/log/absl_check.h"
 #include "absl/strings/substitute.h"
 #include "mediapipe/framework/calculator.pb.h"
 #include "mediapipe/framework/calculator_framework.h"
@@ -16,10 +18,14 @@
 #include "mediapipe/framework/port/opencv_imgcodecs_inc.h"
 #include "mediapipe/framework/port/opencv_imgproc_inc.h"
 #include "mediapipe/framework/port/parse_text_proto.h"
+#include "mediapipe/gpu/multi_pool.h"
 #include "testing/base/public/gmock.h"
 #include "testing/base/public/googletest.h"
+#include "testing/base/public/gunit.h"
 #include "third_party/OpenCV/core.hpp"  // IWYU pragma: keep
+#include "third_party/OpenCV/core/base.hpp"
 #include "third_party/OpenCV/core/mat.hpp"
+#include "third_party/OpenCV/core/types.hpp"
 
 namespace mediapipe {
 
@@ -76,11 +82,12 @@ TEST(ImageTransformationCalculatorTest, NearestNeighborResizing) {
           ->Tag("OUTPUT_DIMENSIONS")
           .packets.push_back(input_output_dim_packet.At(Timestamp(0)));
 
-      MP_ASSERT_OK(runner.Run());
+      ABSL_QCHECK_OK(runner.Run());
       const auto& outputs = runner.Outputs();
-      ASSERT_EQ(outputs.NumEntries(), 1);
+      ABSL_QCHECK_EQ(outputs.NumEntries(), 1);
       const std::vector<Packet>& packets = outputs.Tag("IMAGE").packets;
-      ASSERT_EQ(packets.size(), 1);
+      ABSL_QCHECK_EQ(packets.size(), 1);
+
       const auto& result = packets[0].Get<ImageFrame>();
       ASSERT_EQ(output_dim.first, result.Width());
       ASSERT_EQ(output_dim.second, result.Height());
@@ -137,11 +144,12 @@ TEST(ImageTransformationCalculatorTest,
           ->Tag("OUTPUT_DIMENSIONS")
           .packets.push_back(input_output_dim_packet.At(Timestamp(0)));
 
-      MP_ASSERT_OK(runner.Run());
+      ABSL_QCHECK_OK(runner.Run());
       const auto& outputs = runner.Outputs();
-      ASSERT_EQ(outputs.NumEntries(), 1);
+      ABSL_QCHECK_EQ(outputs.NumEntries(), 1);
       const std::vector<Packet>& packets = outputs.Tag("IMAGE").packets;
-      ASSERT_EQ(packets.size(), 1);
+      ABSL_QCHECK_EQ(packets.size(), 1);
+
       const auto& result = packets[0].Get<ImageFrame>();
       ASSERT_EQ(output_dim.first, result.Width());
       ASSERT_EQ(output_dim.second, result.Height());
@@ -207,17 +215,17 @@ TEST(ImageTransformationCalculatorTest, NearestNeighborResizingGpu) {
       tool::AddVectorSink("output_image", &graph_config, &output_image_packets);
 
       CalculatorGraph graph(graph_config);
-      MP_ASSERT_OK(graph.StartRun({}));
+      ABSL_QCHECK_OK(graph.StartRun({}));
 
-      MP_ASSERT_OK(graph.AddPacketToInputStream(
+      ABSL_QCHECK_OK(graph.AddPacketToInputStream(
           "input_image",
           MakePacket<ImageFrame>(std::move(input_image)).At(Timestamp(0))));
-      MP_ASSERT_OK(graph.AddPacketToInputStream(
+      ABSL_QCHECK_OK(graph.AddPacketToInputStream(
           "image_size",
           MakePacket<std::pair<int, int>>(output_dim).At(Timestamp(0))));
 
-      MP_ASSERT_OK(graph.WaitUntilIdle());
-      ASSERT_THAT(output_image_packets, testing::SizeIs(1));
+      ABSL_QCHECK_OK(graph.WaitUntilIdle());
+      ABSL_QCHECK_EQ(output_image_packets.size(), 1);
 
       const auto& output_image = output_image_packets[0].Get<ImageFrame>();
       ASSERT_EQ(output_dim.first, output_image.Width());
@@ -287,16 +295,16 @@ TEST(ImageTransformationCalculatorTest,
       tool::AddVectorSink("output_image", &graph_config, &output_image_packets);
 
       CalculatorGraph graph(graph_config);
-      MP_ASSERT_OK(graph.StartRun({}));
+      ABSL_QCHECK_OK(graph.StartRun({}));
 
-      MP_ASSERT_OK(graph.AddPacketToInputStream(
+      ABSL_QCHECK_OK(graph.AddPacketToInputStream(
           "input_image", input_image_packet.At(Timestamp(0))));
-      MP_ASSERT_OK(graph.AddPacketToInputStream(
+      ABSL_QCHECK_OK(graph.AddPacketToInputStream(
           "image_size",
           MakePacket<std::pair<int, int>>(output_dim).At(Timestamp(0))));
 
-      MP_ASSERT_OK(graph.WaitUntilIdle());
-      ASSERT_THAT(output_image_packets, testing::SizeIs(1));
+      ABSL_QCHECK_OK(graph.WaitUntilIdle());
+      ABSL_QCHECK_EQ(output_image_packets.size(), 1);
 
       const auto& output_image = output_image_packets[0].Get<ImageFrame>();
       ASSERT_EQ(output_dim.first, output_image.Width());
@@ -311,5 +319,112 @@ TEST(ImageTransformationCalculatorTest,
   }
 }
 
+TEST(ImageTransformationCalculatorTest, FitScalingClearsBackground) {
+  // Regression test for not clearing the background in FIT scaling mode.
+  // First scale an all-red (=r) image from 8x4 to 8x4, so it's a plain copy:
+  //   rrrrrrrr
+  //   rrrrrrrr
+  //   rrrrrrrr
+  //   rrrrrrrr
+  // Then scale an all-blue image from 4x4 to 8x4 in FIT mode. This should
+  // introduce dark yellow (=y) letterboxes left and right due to padding_color:
+  //   yybbbbyy
+  //   yybbbbyy
+  //   yybbbbyy
+  //   yybbbbyy
+  // We make sure that the all-red buffer gets reused. Without clearing the
+  // background, the blue (=b) image will have red letterboxes:
+  //   rrbbbbrr
+  //   rrbbbbrr
+  //   rrbbbbrr
+  //   rrbbbbrr
+
+  constexpr int kSmall = 4, kLarge = 8;
+  ImageFrame input_image_red(ImageFormat::SRGBA, kLarge, kSmall);
+  cv::Mat input_image_red_mat = formats::MatView(&input_image_red);
+  input_image_red_mat = cv::Scalar(255, 0, 0, 255);
+
+  ImageFrame input_image_blue(ImageFormat::SRGBA, kSmall, kSmall);
+  cv::Mat input_image_blue_mat = formats::MatView(&input_image_blue);
+  input_image_blue_mat = cv::Scalar(0, 0, 255, 255);
+
+  Packet input_image_red_packet =
+      MakePacket<ImageFrame>(std::move(input_image_red));
+  Packet input_image_blue_packet =
+      MakePacket<ImageFrame>(std::move(input_image_blue));
+
+  CalculatorGraphConfig graph_config =
+      ParseTextProtoOrDie<CalculatorGraphConfig>(absl::Substitute(
+          R"pb(
+            input_stream: "input_image"
+            output_stream: "output_image"
+
+            node {
+              calculator: "ImageFrameToGpuBufferCalculator"
+              input_stream: "input_image"
+              output_stream: "input_image_gpu"
+            }
+
+            node {
+              calculator: "ImageTransformationCalculator"
+              input_stream: "IMAGE_GPU:input_image_gpu"
+              output_stream: "IMAGE_GPU:output_image_gpu"
+              options: {
+                [mediapipe.ImageTransformationCalculatorOptions.ext]: {
+                  scale_mode: FIT
+                  output_width: $0,
+                  output_height: $1,
+                  padding_color: { red: 128, green: 128, blue: 0 }
+                }
+              }
+            }
+
+            node {
+              calculator: "GpuBufferToImageFrameCalculator"
+              input_stream: "output_image_gpu"
+              output_stream: "output_image"
+            })pb",
+          kLarge, kSmall));
+
+  std::vector<Packet> output_image_packets;
+  tool::AddVectorSink("output_image", &graph_config, &output_image_packets);
+
+  CalculatorGraph graph(graph_config);
+  ABSL_QCHECK_OK(graph.StartRun({}));
+
+  // Send the red image multiple times to cause the GPU pool to actually use
+  // a pool.
+  int num_red_packets =
+      std::max(kDefaultMultiPoolOptions.min_requests_before_pool, 1);
+  for (int n = 0; n < num_red_packets; ++n) {
+    ABSL_QCHECK_OK(graph.AddPacketToInputStream(
+        "input_image", input_image_red_packet.At(Timestamp(n))));
+  }
+  ABSL_QCHECK_OK(graph.AddPacketToInputStream(
+      "input_image", input_image_blue_packet.At(Timestamp(num_red_packets))));
+
+  ABSL_QCHECK_OK(graph.WaitUntilIdle());
+  ABSL_QCHECK_EQ(output_image_packets.size(), num_red_packets + 1);
+
+  const auto& output_image_red = output_image_packets[0].Get<ImageFrame>();
+  const auto& output_image_blue =
+      output_image_packets[num_red_packets].Get<ImageFrame>();
+
+  ABSL_QCHECK_EQ(output_image_red.Width(), kLarge);
+  ABSL_QCHECK_EQ(output_image_red.Height(), kSmall);
+  ABSL_QCHECK_EQ(output_image_blue.Width(), kLarge);
+  ABSL_QCHECK_EQ(output_image_blue.Height(), kSmall);
+
+  cv::Mat output_image_blue_mat = formats::MatView(&output_image_blue);
+  ImageFrame expected_image_blue(ImageFormat::SRGBA, kLarge, kSmall);
+  cv::Mat expected_image_blue_mat = formats::MatView(&expected_image_blue);
+  expected_image_blue_mat = cv::Scalar(128, 128, 0, 255);
+  cv::Rect rect((kLarge - kSmall) / 2, 0, kSmall, kSmall);
+  cv::rectangle(expected_image_blue_mat, rect, cv::Scalar(0, 0, 255, 255),
+                cv::FILLED);
+  EXPECT_EQ(cv::sum(cv::sum(output_image_blue_mat != expected_image_blue_mat)),
+            cv::Scalar(0));
+}
+
 }  // namespace
 }  // namespace mediapipe

mediapipe/examples/desktop/autoflip/quality/kinematic_path_solver.h

@@ -43,9 +43,9 @@ class KinematicPathSolver {
         initialized_(false),
         pixels_per_degree_(pixels_per_degree) {}
   // Add an observation (detection) at a position and time.
-  absl::Status AddObservation(int position, const uint64 time_us);
+  absl::Status AddObservation(int position, const uint64_t time_us);
   // Get the predicted position at a time.
-  absl::Status UpdatePrediction(const int64 time_us);
+  absl::Status UpdatePrediction(const int64_t time_us);
   // Get the state at a time, as an int.
   absl::Status GetState(int* position);
   // Get the state at a time, as a float.
@@ -63,7 +63,7 @@ class KinematicPathSolver {
   bool IsMotionTooSmall(double delta_degs);
   // Check if a position measurement will cause the camera to be in motion
   // without updating the internal state.
-  absl::Status PredictMotionState(int position, const uint64 time_us,
+  absl::Status PredictMotionState(int position, const uint64_t time_us,
                                   bool* state);
   // Clear any history buffer of positions that are used when
   // filtering_time_window_us is set to a non-zero value.
@@ -85,9 +85,9 @@ class KinematicPathSolver {
   double current_position_px_;
   double prior_position_px_;
   double current_velocity_deg_per_s_;
-  uint64 current_time_ = 0;
+  uint64_t current_time_ = 0;
   // History of observations (second) and their time (first).
-  std::deque<std::pair<uint64, int>> raw_positions_at_time_;
+  std::deque<std::pair<uint64_t, int>> raw_positions_at_time_;
   // Current target position.
   double target_position_px_;
   // Defines if the camera is moving to a target (true) or reached a target

mediapipe/examples/desktop/autoflip/quality/scene_camera_motion_analyzer.h

@@ -67,7 +67,7 @@ class SceneCameraMotionAnalyzer {
       const KeyFrameCropOptions& key_frame_crop_options,
       const std::vector<KeyFrameCropResult>& key_frame_crop_results,
       const int scene_frame_width, const int scene_frame_height,
-      const std::vector<int64>& scene_frame_timestamps,
+      const std::vector<int64_t>& scene_frame_timestamps,
       const bool has_solid_color_background,
       SceneKeyFrameCropSummary* scene_summary,
       std::vector<FocusPointFrame>* focus_point_frames,
@@ -78,7 +78,7 @@ class SceneCameraMotionAnalyzer {
   // crop window in SceneKeyFrameCropSummary in the case of steady motion.
   absl::Status DecideCameraMotionType(
       const KeyFrameCropOptions& key_frame_crop_options,
-      const double scene_span_sec, const int64 end_time_us,
+      const double scene_span_sec, const int64_t end_time_us,
       SceneKeyFrameCropSummary* scene_summary,
       SceneCameraMotion* scene_camera_motion) const;
 
@@ -87,7 +87,7 @@ class SceneCameraMotionAnalyzer {
   absl::Status PopulateFocusPointFrames(
       const SceneKeyFrameCropSummary& scene_summary,
       const SceneCameraMotion& scene_camera_motion,
-      const std::vector<int64>& scene_frame_timestamps,
+      const std::vector<int64_t>& scene_frame_timestamps,
       std::vector<FocusPointFrame>* focus_point_frames) const;
 
  private:
@@ -118,7 +118,7 @@ class SceneCameraMotionAnalyzer {
   absl::Status PopulateFocusPointFramesForTracking(
       const SceneKeyFrameCropSummary& scene_summary,
       const FocusPointFrameType focus_point_frame_type,
-      const std::vector<int64>& scene_frame_timestamps,
+      const std::vector<int64_t>& scene_frame_timestamps,
       std::vector<FocusPointFrame>* focus_point_frames) const;
 
   // Decide to use steady motion.
@@ -142,7 +142,7 @@ class SceneCameraMotionAnalyzer {
 
   // Last position
   SceneCameraMotion last_scene_with_salient_region_;
-  int64 time_since_last_salient_region_us_;
+  int64_t time_since_last_salient_region_us_;
 
   // Scene has solid color background.
   bool has_solid_color_background_;

mediapipe/examples/desktop/autoflip/quality/scene_cropper.h

@@ -62,7 +62,7 @@ class SceneCropper {
   // TODO: split this function into two separate functions.
   absl::Status CropFrames(
       const SceneKeyFrameCropSummary& scene_summary,
-      const std::vector<int64>& scene_timestamps,
+      const std::vector<int64_t>& scene_timestamps,
       const std::vector<bool>& is_key_frames,
       const std::vector<cv::Mat>& scene_frames_or_empty,
       const std::vector<FocusPointFrame>& focus_point_frames,
@@ -73,7 +73,7 @@ class SceneCropper {
 
   absl::Status ProcessKinematicPathSolver(
       const SceneKeyFrameCropSummary& scene_summary,
-      const std::vector<int64>& scene_timestamps,
+      const std::vector<int64_t>& scene_timestamps,
       const std::vector<bool>& is_key_frames,
       const std::vector<FocusPointFrame>& focus_point_frames,
       const bool continue_last_scene, std::vector<cv::Mat>* all_xforms);

mediapipe/examples/desktop/autoflip/quality/utils.h

@@ -29,7 +29,7 @@ namespace autoflip {
 // Packs detected features and timestamp (ms) into a KeyFrameInfo object. Scales
 // features back to the original frame size if features have been detected on a
 // different frame size.
-absl::Status PackKeyFrameInfo(const int64 frame_timestamp_ms,
+absl::Status PackKeyFrameInfo(const int64_t frame_timestamp_ms,
                               const DetectionSet& detections,
                               const int original_frame_width,
                               const int original_frame_height,
@@ -71,7 +71,7 @@ absl::Status ComputeSceneStaticBordersSize(
 // interpolation functions in Lab space using input timestamps.
 absl::Status FindSolidBackgroundColor(
     const std::vector<StaticFeatures>& static_features,
-    const std::vector<int64>& static_features_timestamps,
+    const std::vector<int64_t>& static_features_timestamps,
     const double min_fraction_solid_background_color,
     bool* has_solid_background,
     PiecewiseLinearFunction* background_color_l_function,

mediapipe/framework/formats/BUILD

@@ -155,6 +155,27 @@ cc_library(
     ],
 )
 
+cc_library(
+    name = "hardware_buffer",
+    srcs = ["hardware_buffer_android.cc"],
+    hdrs = ["hardware_buffer.h"],
+    linkopts = select({
+        "//conditions:default": [],
+        # Option for vendor binaries to avoid linking libandroid.so.
+        "//mediapipe/framework:android_no_jni": [],
+        "//mediapipe:android": ["-landroid"],
+        ":android_link_native_window": [
+            "-lnativewindow",  # Provides <android/hardware_buffer.h> to vendor binaries on Android API >= 26.
+        ],
+    }),
+    visibility = ["//visibility:private"],
+    deps = [
+        "//mediapipe/framework/port:ret_check",
+        "//mediapipe/framework/port:statusor",
+        "@com_google_absl//absl/log:absl_check",
+    ],
+)
+
 cc_library(
     name = "image_frame",
     srcs = ["image_frame.cc"],
@@ -493,10 +514,6 @@ cc_library(
         "//conditions:default": [],
         # Option for vendor binaries to avoid linking libandroid.so.
         "//mediapipe/framework:android_no_jni": [],
-        "//mediapipe:android": ["-landroid"],
-        ":android_link_native_window": [
-            "-lnativewindow",  # Provides <android/hardware_buffer.h> to vendor binaries on Android API >= 26.
-        ],
     }),
     deps = [
         "//mediapipe/framework:port",
@@ -511,9 +528,16 @@ cc_library(
             "//mediapipe/gpu:gl_base",
             "//mediapipe/gpu:gl_context",
         ],
-           }) +
-           select({
-               "//conditions:default": [],
+        "//mediapipe:android": [
+            ":hardware_buffer",
+            "//mediapipe/gpu:gl_base",
+            "//mediapipe/gpu:gl_context",
+        ],
+        ":android_link_native_window": [
+            ":hardware_buffer",
+            "//mediapipe/gpu:gl_base",
+            "//mediapipe/gpu:gl_context",
+        ],
     }),
 )
 

mediapipe/framework/formats/deleting_file.cc

@@ -17,10 +17,25 @@
 
 #include <stdio.h>
 
+#include <utility>
+
 #include "absl/log/absl_log.h"
 
 namespace mediapipe {
 
+DeletingFile::DeletingFile(DeletingFile&& other)
+    : path_(std::move(other.path_)),
+      delete_on_destruction_(other.delete_on_destruction_) {
+  other.delete_on_destruction_ = false;
+}
+
+DeletingFile& DeletingFile::operator=(DeletingFile&& other) {
+  path_ = std::move(other.path_);
+  delete_on_destruction_ = other.delete_on_destruction_;
+  other.delete_on_destruction_ = false;
+  return *this;
+}
+
 DeletingFile::DeletingFile(const std::string& path, bool delete_on_destruction)
     : path_(path), delete_on_destruction_(delete_on_destruction) {}
 

mediapipe/framework/formats/deleting_file.h

@@ -28,6 +28,11 @@ class DeletingFile {
   DeletingFile(const DeletingFile&) = delete;
   DeletingFile& operator=(const DeletingFile&) = delete;
 
+  // DeletingFile is movable. The moved-from object remains in valid but
+  // unspecified state and will not perform any operations on destruction.
+  DeletingFile(DeletingFile&& other);
+  DeletingFile& operator=(DeletingFile&& other);
+
   // Provide the path to the file and whether the file should be deleted
   // when this object is destroyed.
   DeletingFile(const std::string& path, bool delete_on_destruction);

mediapipe/framework/formats/frame_buffer.h

@@ -16,11 +16,11 @@ limitations under the License.
 #ifndef MEDIAPIPE_FRAMEWORK_FORMATS_FRAME_BUFFER_H_
 #define MEDIAPIPE_FRAMEWORK_FORMATS_FRAME_BUFFER_H_
 
+#include <cstdint>
 #include <vector>
 
 #include "absl/log/absl_check.h"
 #include "absl/status/statusor.h"
-#include "mediapipe/framework/port/integral_types.h"
 
 namespace mediapipe {
 
@@ -76,13 +76,13 @@ class FrameBuffer {
 
   // Plane encapsulates buffer and stride information.
   struct Plane {
-    Plane(uint8* buffer, Stride stride) : buffer_(buffer), stride_(stride) {}
-    const uint8* buffer() const { return buffer_; }
-    uint8* mutable_buffer() { return buffer_; }
+    Plane(uint8_t* buffer, Stride stride) : buffer_(buffer), stride_(stride) {}
+    const uint8_t* buffer() const { return buffer_; }
+    uint8_t* mutable_buffer() { return buffer_; }
     Stride stride() const { return stride_; }
 
    private:
-    uint8* buffer_;
+    uint8_t* buffer_;
     Stride stride_;
   };
 
@@ -121,9 +121,9 @@ class FrameBuffer {
 
   // YUV data structure.
   struct YuvData {
-    const uint8* y_buffer;
-    const uint8* u_buffer;
-    const uint8* v_buffer;
+    const uint8_t* y_buffer;
+    const uint8_t* u_buffer;
+    const uint8_t* v_buffer;
     // Y buffer row stride in bytes.
     int y_row_stride;
     // U/V buffer row stride in bytes.

mediapipe/framework/formats/hardware_buffer.h

@@ -0,0 +1,167 @@
+// Copyright 2019 The MediaPipe Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#ifndef MEDIAPIPE_FRAMEWORK_FORMATS_HARDWARE_BUFFER_H_
+#define MEDIAPIPE_FRAMEWORK_FORMATS_HARDWARE_BUFFER_H_
+
+#include <cstdint>
+#include <memory>
+#include <optional>
+#include <utility>
+
+#include "absl/base/attributes.h"
+#include "absl/status/status.h"
+#include "absl/status/statusor.h"
+
+typedef struct AHardwareBuffer AHardwareBuffer;
+
+namespace mediapipe {
+
+struct HardwareBufferSpec {
+  // Buffer pixel formats. See NDK's hardware_buffer.h for descriptions.
+  enum {
+    // This must be kept in sync with NDK's hardware_buffer.h
+    AHARDWAREBUFFER_FORMAT_R8G8B8A8_UNORM = 0x01,
+    AHARDWAREBUFFER_FORMAT_R8G8B8_UNORM = 0x03,
+    AHARDWAREBUFFER_FORMAT_R16G16B16A16_FLOAT = 0x16,
+    AHARDWAREBUFFER_FORMAT_BLOB = 0x21,
+    AHARDWAREBUFFER_FORMAT_R8_UNORM = 0x38,
+  };
+
+  // Buffer usage descriptions. See NDK's hardware_buffer.h for descriptions.
+  enum {
+    // This must be kept in sync with NDK's hardware_buffer.h
+    AHARDWAREBUFFER_USAGE_CPU_READ_NEVER = 0x0UL,
+    AHARDWAREBUFFER_USAGE_CPU_READ_RARELY = 0x2UL,
+    AHARDWAREBUFFER_USAGE_CPU_READ_OFTEN = 0x3UL,
+    AHARDWAREBUFFER_USAGE_CPU_WRITE_NEVER = UINT64_C(0) << 4,
+    AHARDWAREBUFFER_USAGE_CPU_WRITE_RARELY = UINT64_C(2) << 4,
+    AHARDWAREBUFFER_USAGE_CPU_WRITE_OFTEN = UINT64_C(3) << 4,
+    AHARDWAREBUFFER_USAGE_GPU_SAMPLED_IMAGE = UINT64_C(1) << 8,
+    AHARDWAREBUFFER_USAGE_GPU_FRAMEBUFFER = UINT64_C(1) << 9,
+    AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER = UINT64_C(1) << 24,
+  };
+
+  // Hashing required to use HardwareBufferSpec as key in buffer pools. See
+  // absl::Hash for details.
+  template <typename H>
+  friend H AbslHashValue(H h, const HardwareBufferSpec& spec) {
+    return H::combine(std::move(h), spec.width, spec.height, spec.layers,
+                      spec.format, spec.usage);
+  }
+
+  uint32_t width = 0;
+  uint32_t height = 0;
+  uint32_t layers = 0;
+  uint32_t format = 0;
+  uint64_t usage = 0;
+};
+
+// Equality operators
+inline bool operator==(const HardwareBufferSpec& lhs,
+                       const HardwareBufferSpec& rhs) {
+  return lhs.width == rhs.width && lhs.height == rhs.height &&
+         lhs.layers == rhs.layers && lhs.format == rhs.format &&
+         lhs.usage == rhs.usage;
+}
+inline bool operator!=(const HardwareBufferSpec& lhs,
+                       const HardwareBufferSpec& rhs) {
+  return !operator==(lhs, rhs);
+}
+
+// For internal use only. Thinly wraps the Android NDK AHardwareBuffer.
+class HardwareBuffer {
+ public:
+  // Constructs a HardwareBuffer instance from a newly allocated Android NDK
+  // AHardwareBuffer.
+  static absl::StatusOr<HardwareBuffer> Create(const HardwareBufferSpec& spec);
+
+  // Destructs the HardwareBuffer, releasing the AHardwareBuffer.
+  ~HardwareBuffer();
+
+  // Support HardwareBuffer moves.
+  HardwareBuffer(HardwareBuffer&& other);
+
+  // Delete assignment and copy constructors.
+  HardwareBuffer(HardwareBuffer& other) = delete;
+  HardwareBuffer(const HardwareBuffer& other) = delete;
+  HardwareBuffer& operator=(const HardwareBuffer&) = delete;
+
+  // Returns true if AHWB is supported.
+  static bool IsSupported();
+
+  // Lock the hardware buffer for the given usage flags. fence_file_descriptor
+  // specifies a fence file descriptor on which to wait before locking the
+  // buffer. Returns raw memory address if lock is successful, nullptr
+  // otherwise.
+  ABSL_MUST_USE_RESULT absl::StatusOr<void*> Lock(
+      uint64_t usage, std::optional<int> fence_file_descriptor = std::nullopt);
+
+  // Unlocks the hardware buffer synchronously. This method blocks until
+  // unlocking is complete.
+  absl::Status Unlock();
+
+  // Unlocks the hardware buffer asynchronously. It returns a file_descriptor
+  // which can be used as a fence that is signaled once unlocking is complete.
+  absl::StatusOr<int> UnlockAsync();
+
+  // Returns the underlying raw AHardwareBuffer pointer to be used directly with
+  // AHardwareBuffer APIs.
+  AHardwareBuffer* GetAHardwareBuffer() const { return ahw_buffer_; }
+
+  // Returns whether this HardwareBuffer contains a valid AHardwareBuffer.
+  bool IsValid() const { return ahw_buffer_ != nullptr; }
+
+  // Returns whether this HardwareBuffer is locked.
+  bool IsLocked() const { return is_locked_; }
+
+  // Releases the AHardwareBuffer.
+  void Reset();
+
+  // Ahwb's are aligned to an implementation specific cacheline size.
+  uint32_t GetAlignedWidth() const;
+
+  // Returns buffer spec.
+  const HardwareBufferSpec& spec() const { return spec_; }
+
+ private:
+  // Allocates an AHardwareBuffer instance;
+  static absl::StatusOr<AHardwareBuffer*> AllocateAHardwareBuffer(
+      const HardwareBufferSpec& spec);
+
+  // Constructs a HardwareBuffer instance from an already aquired
+  // AHardwareBuffer instance and its spec.
+  HardwareBuffer(const HardwareBufferSpec& spec, AHardwareBuffer* ahwb);
+
+  // Unlocks the hardware buffer. If fence_file_descriptor_ptr is not nullptr,
+  // the function won't block and instead fence_file_descriptor_ptr will be set
+  // to a file descriptor to become signaled once unlocking is complete.
+  absl::Status UnlockInternal(int* fence_file_descriptor_ptr);
+
+  // Releases ahw_buffer_ AHardwareBuffer instance;
+  absl::Status ReleaseAHardwareBuffer();
+
+  // Buffer spec.
+  HardwareBufferSpec spec_ = {};
+
+  // Android NDK AHardwareBuffer.
+  AHardwareBuffer* ahw_buffer_ = nullptr;
+
+  // Indicates if AHardwareBuffer is locked for reading or writing.
+  bool is_locked_ = false;
+};
+
+}  // namespace mediapipe
+
+#endif  // MEDIAPIPE_FRAMEWORK_FORMATS_AHWB_BUFFER_H_

mediapipe/framework/formats/hardware_buffer_android.cc

@@ -0,0 +1,152 @@
+// Copyright 2019 The MediaPipe Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#if !defined(MEDIAPIPE_NO_JNI) && \
+    (__ANDROID_API__ >= 26 ||     \
+     defined(__ANDROID_UNAVAILABLE_SYMBOLS_ARE_WEAK__))
+
+#include <android/hardware_buffer.h>
+
+#include <memory>
+
+#include "absl/log/absl_check.h"
+#include "absl/status/status.h"
+#include "absl/status/statusor.h"
+#include "mediapipe/framework/formats/hardware_buffer.h"
+#include "mediapipe/framework/port/ret_check.h"
+
+namespace mediapipe {
+
+HardwareBuffer::HardwareBuffer(HardwareBuffer &&other) {
+  spec_ = std::exchange(other.spec_, {});
+  ahw_buffer_ = std::exchange(other.ahw_buffer_, nullptr);
+  is_locked_ = std::exchange(other.is_locked_, false);
+}
+
+HardwareBuffer::HardwareBuffer(const HardwareBufferSpec &spec,
+                               AHardwareBuffer *ahwb)
+    : spec_(spec), ahw_buffer_(ahwb), is_locked_(false) {}
+
+HardwareBuffer::~HardwareBuffer() { Reset(); }
+
+absl::StatusOr<HardwareBuffer> HardwareBuffer::Create(
+    const HardwareBufferSpec &spec) {
+  MP_ASSIGN_OR_RETURN(AHardwareBuffer * ahwb, AllocateAHardwareBuffer(spec));
+  return HardwareBuffer(spec, ahwb);
+}
+
+bool HardwareBuffer::IsSupported() {
+  if (__builtin_available(android 26, *)) {
+    return true;
+  }
+  return false;
+}
+
+absl::StatusOr<AHardwareBuffer *> HardwareBuffer::AllocateAHardwareBuffer(
+    const HardwareBufferSpec &spec) {
+  RET_CHECK(IsSupported()) << "AndroidHWBuffers not supported";
+
+  AHardwareBuffer *output = nullptr;
+  int error = 0;
+  if (__builtin_available(android 26, *)) {
+    AHardwareBuffer_Desc desc = {
+        .width = spec.width,
+        .height = spec.height,
+        .layers = spec.layers,
+        .format = spec.format,
+        .usage = spec.usage,
+    };
+    error = AHardwareBuffer_allocate(&desc, &output);
+  }
+  RET_CHECK(!error && output != nullptr) << "AHardwareBuffer_allocate failed";
+  return output;
+}
+
+absl::Status HardwareBuffer::ReleaseAHardwareBuffer() {
+  if (ahw_buffer_ == nullptr) {
+    return absl::OkStatus();
+  }
+  if (is_locked_) {
+    MP_RETURN_IF_ERROR(Unlock());
+  }
+  if (__builtin_available(android 26, *)) {
+    AHardwareBuffer_release(ahw_buffer_);
+  }
+  spec_ = {};
+  ahw_buffer_ = nullptr;
+  return absl::OkStatus();
+}
+
+absl::StatusOr<void *> HardwareBuffer::Lock(
+    uint64_t usage, std::optional<int> fence_file_descriptor) {
+  RET_CHECK(ahw_buffer_ != nullptr) << "Hardware Buffer not allocated";
+  RET_CHECK(!is_locked_) << "Hardware Buffer already locked";
+  void *mem = nullptr;
+  if (__builtin_available(android 26, *)) {
+    const int error = AHardwareBuffer_lock(
+        ahw_buffer_, usage,
+        fence_file_descriptor.has_value() ? *fence_file_descriptor : -1,
+        nullptr, &mem);
+    RET_CHECK(error == 0) << "Hardware Buffer lock failed. Error: " << error;
+  }
+  is_locked_ = true;
+  return mem;
+}
+
+absl::Status HardwareBuffer::Unlock() {
+  return UnlockInternal(/*fence_file_descriptor=*/nullptr);
+}
+
+absl::StatusOr<int> HardwareBuffer::UnlockAsync() {
+  int fence_file_descriptor = -1;
+  MP_RETURN_IF_ERROR(UnlockInternal(&fence_file_descriptor));
+  return fence_file_descriptor;
+}
+
+absl::Status HardwareBuffer::UnlockInternal(int *fence_file_descriptor) {
+  RET_CHECK(ahw_buffer_ != nullptr) << "Hardware Buffer not allocated";
+  if (!is_locked_) {
+    return absl::OkStatus();
+  }
+  if (__builtin_available(android 26, *)) {
+    const int error =
+        AHardwareBuffer_unlock(ahw_buffer_, fence_file_descriptor);
+    RET_CHECK(error == 0) << "Hardware Buffer unlock failed. error: " << error;
+  }
+  is_locked_ = false;
+  return absl::OkStatus();
+}
+
+uint32_t HardwareBuffer::GetAlignedWidth() const {
+  if (__builtin_available(android 26, *)) {
+    ABSL_CHECK(ahw_buffer_ != nullptr) << "Hardware Buffer not allocated";
+    AHardwareBuffer_Desc desc = {};
+    AHardwareBuffer_describe(ahw_buffer_, &desc);
+    ABSL_CHECK_GT(desc.stride, 0);
+    return desc.stride;
+  }
+  return 0;
+}
+
+void HardwareBuffer::Reset() {
+  const auto success = ReleaseAHardwareBuffer();
+  if (!success.ok()) {
+    ABSL_LOG(DFATAL) << "Failed to release AHardwareBuffer: " << success;
+  }
+}
+
+}  // namespace mediapipe
+
+#endif  // !defined(MEDIAPIPE_NO_JNI) && (__ANDROID_API__>= 26 ||
+        // defined(__ANDROID_UNAVAILABLE_SYMBOLS_ARE_WEAK__))

mediapipe/framework/formats/hardware_buffer_test.cc

@@ -0,0 +1,131 @@
+#include "mediapipe/framework/formats/hardware_buffer.h"
+
+#include <android/hardware_buffer.h>
+
+#include <memory>
+
+#include "base/logging.h"
+#include "mediapipe/framework/port/status_macros.h"
+#include "testing/base/public/gmock.h"
+#include "testing/base/public/gunit.h"
+
+namespace mediapipe {
+
+namespace {
+
+HardwareBufferSpec GetTestHardwareBufferSpec(uint32_t size_bytes) {
+  return {.width = size_bytes,
+          .height = 1,
+          .layers = 1,
+          .format = HardwareBufferSpec::AHARDWAREBUFFER_FORMAT_BLOB,
+          .usage = HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_WRITE_RARELY |
+                   HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_WRITE_OFTEN |
+                   HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_READ_OFTEN |
+                   HardwareBufferSpec::AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER};
+}
+
+TEST(HardwareBufferTest, ShouldConstructValidAHardwareBuffer) {
+  MP_ASSERT_OK_AND_ASSIGN(
+      HardwareBuffer hardware_buffer,
+      HardwareBuffer::Create(GetTestHardwareBufferSpec(/*size_bytes=*/123)));
+  EXPECT_NE(hardware_buffer.GetAHardwareBuffer(), nullptr);
+  EXPECT_TRUE(hardware_buffer.IsValid());
+}
+
+TEST(HardwareBufferTest, ShouldResetValidAHardwareBuffer) {
+  MP_ASSERT_OK_AND_ASSIGN(
+      HardwareBuffer hardware_buffer,
+      HardwareBuffer::Create(GetTestHardwareBufferSpec(/*size_bytes=*/123)));
+  EXPECT_TRUE(hardware_buffer.IsValid());
+  EXPECT_NE(*hardware_buffer.Lock(
+                HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_WRITE_RARELY),
+            nullptr);
+  EXPECT_TRUE(hardware_buffer.IsLocked());
+
+  hardware_buffer.Reset();
+
+  EXPECT_FALSE(hardware_buffer.IsValid());
+  EXPECT_FALSE(hardware_buffer.IsLocked());
+}
+
+TEST(HardwareBufferTest, ShouldAllocateRequestedBufferSize) {
+  constexpr int kBufferSize = 123;
+  const HardwareBufferSpec spec = GetTestHardwareBufferSpec(kBufferSize);
+  MP_ASSERT_OK_AND_ASSIGN(HardwareBuffer hardware_buffer,
+                          HardwareBuffer::Create(spec));
+
+  EXPECT_TRUE(hardware_buffer.IsValid());
+  if (__builtin_available(android 26, *)) {
+    AHardwareBuffer_Desc desc;
+    AHardwareBuffer_describe(hardware_buffer.GetAHardwareBuffer(), &desc);
+    EXPECT_EQ(desc.width, spec.width);
+    EXPECT_EQ(desc.height, spec.height);
+    EXPECT_EQ(desc.layers, spec.layers);
+    EXPECT_EQ(desc.format, spec.format);
+    EXPECT_EQ(desc.usage, spec.usage);
+  }
+  EXPECT_EQ(hardware_buffer.spec().width, spec.width);
+  EXPECT_EQ(hardware_buffer.spec().height, spec.height);
+  EXPECT_EQ(hardware_buffer.spec().layers, spec.layers);
+  EXPECT_EQ(hardware_buffer.spec().format, spec.format);
+  EXPECT_EQ(hardware_buffer.spec().usage, spec.usage);
+}
+
+TEST(HardwareBufferTest, ShouldSupportMoveConstructor) {
+  constexpr int kBufferSize = 123;
+  const auto spec = GetTestHardwareBufferSpec(kBufferSize);
+  MP_ASSERT_OK_AND_ASSIGN(HardwareBuffer hardware_buffer_a,
+                          HardwareBuffer::Create(spec));
+  EXPECT_TRUE(hardware_buffer_a.IsValid());
+  void* const ahardware_buffer_ptr_a = hardware_buffer_a.GetAHardwareBuffer();
+  EXPECT_NE(ahardware_buffer_ptr_a, nullptr);
+  EXPECT_FALSE(hardware_buffer_a.IsLocked());
+  MP_ASSERT_OK_AND_ASSIGN(
+      void* const hardware_buffer_a_locked_ptr,
+      hardware_buffer_a.Lock(
+          HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_WRITE_RARELY));
+  EXPECT_NE(hardware_buffer_a_locked_ptr, nullptr);
+  EXPECT_TRUE(hardware_buffer_a.IsLocked());
+
+  HardwareBuffer hardware_buffer_b(std::move(hardware_buffer_a));
+
+  EXPECT_FALSE(hardware_buffer_a.IsValid());
+  EXPECT_FALSE(hardware_buffer_a.IsLocked());
+  void* const ahardware_buffer_ptr_b = hardware_buffer_b.GetAHardwareBuffer();
+  EXPECT_EQ(ahardware_buffer_ptr_a, ahardware_buffer_ptr_b);
+  EXPECT_TRUE(hardware_buffer_b.IsValid());
+  EXPECT_TRUE(hardware_buffer_b.IsLocked());
+
+  EXPECT_EQ(hardware_buffer_a.spec(), HardwareBufferSpec());
+  EXPECT_EQ(hardware_buffer_b.spec(), spec);
+
+  MP_ASSERT_OK(hardware_buffer_b.Unlock());
+}
+
+TEST(HardwareBufferTest, ShouldSupportReadWrite) {
+  constexpr std::string_view kTestString = "TestString";
+  constexpr int kBufferSize = kTestString.size();
+  MP_ASSERT_OK_AND_ASSIGN(
+      HardwareBuffer hardware_buffer,
+      HardwareBuffer::Create(GetTestHardwareBufferSpec(kBufferSize)));
+
+  // Write test string.
+  MP_ASSERT_OK_AND_ASSIGN(
+      void* const write_ptr,
+      hardware_buffer.Lock(
+          HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_WRITE_RARELY));
+  memcpy(write_ptr, kTestString.data(), kBufferSize);
+  MP_ASSERT_OK(hardware_buffer.Unlock());
+
+  // Read test string.
+  MP_ASSERT_OK_AND_ASSIGN(
+      void* const read_ptr,
+      hardware_buffer.Lock(
+          HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_READ_RARELY));
+  EXPECT_EQ(memcmp(read_ptr, kTestString.data(), kBufferSize), 0);
+  MP_ASSERT_OK(hardware_buffer.Unlock());
+}
+
+}  // namespace
+
+}  // namespace mediapipe

mediapipe/framework/formats/image.h

@@ -15,6 +15,7 @@
 #ifndef MEDIAPIPE_FRAMEWORK_FORMATS_IMAGE_H_
 #define MEDIAPIPE_FRAMEWORK_FORMATS_IMAGE_H_
 
+#include <cstdint>
 #include <utility>
 
 #include "absl/synchronization/mutex.h"
@@ -207,7 +208,7 @@ inline void Image::UnlockPixels() const {}
 //   Image buf = ...
 //   {
 //     PixelLock lock(&buf);
-//     uint8* buf_ptr = lock.Pixels();
+//     uint8_t* buf_ptr = lock.Pixels();
 //     ... use buf_ptr to access pixel data ...
 //     ... lock released automatically at end of scope ...
 //   }
@@ -228,7 +229,7 @@ class PixelReadLock {
   }
   PixelReadLock(const PixelReadLock&) = delete;
 
-  const uint8* Pixels() const {
+  const uint8_t* Pixels() const {
     if (frame_) return frame_->PixelData();
     return nullptr;
   }
@@ -254,7 +255,7 @@ class PixelWriteLock {
   }
   PixelWriteLock(const PixelWriteLock&) = delete;
 
-  uint8* Pixels() {
+  uint8_t* Pixels() {
     if (frame_) return frame_->MutablePixelData();
     return nullptr;
   }

mediapipe/framework/formats/image_frame.h

@@ -35,13 +35,13 @@
 #ifndef MEDIAPIPE_FRAMEWORK_FORMATS_IMAGE_FRAME_H_
 #define MEDIAPIPE_FRAMEWORK_FORMATS_IMAGE_FRAME_H_
 
+#include <cstdint>
 #include <functional>
 #include <memory>
 #include <string>
 
 #include "mediapipe/framework/formats/image_format.pb.h"
 #include "mediapipe/framework/port.h"
-#include "mediapipe/framework/port/integral_types.h"
 #include "mediapipe/framework/tool/type_util.h"
 
 #define IMAGE_FRAME_RAW_IMAGE MEDIAPIPE_HAS_RTTI
@@ -63,7 +63,7 @@ namespace mediapipe {
 // stored with row padding for alignment purposes.
 class ImageFrame {
  public:
-  typedef std::function<void(uint8*)> Deleter;
+  typedef std::function<void(uint8_t*)> Deleter;
 
   // This class offers a few standard delete functions and retains
   // compatibility with the previous API.
@@ -78,13 +78,13 @@ class ImageFrame {
   // Use a default alignment boundary of 16 because Intel SSE2 instructions may
   // incur performance penalty when accessing data not aligned on a 16-byte
   // boundary. FFmpeg requires at least this level of alignment.
-  static const uint32 kDefaultAlignmentBoundary = 16;
+  static const uint32_t kDefaultAlignmentBoundary = 16;
 
   // If the pixel data of an ImageFrame will be passed to an OpenGL function
   // such as glTexImage2D() or glReadPixels(), use a four-byte alignment
   // boundary because that is the initial value of the OpenGL GL_PACK_ALIGNMENT
   // and GL_UNPACK_ALIGNMENT parameters.
-  static const uint32 kGlDefaultAlignmentBoundary = 4;
+  static const uint32_t kGlDefaultAlignmentBoundary = 4;
 
   // Returns number of channels for an ImageFormat.
   static int NumberOfChannelsForFormat(ImageFormat::Format format);
@@ -104,7 +104,7 @@ class ImageFrame {
   // must be a power of 2 (the number 1 is valid, and means the data will
   // be stored contiguously).
   ImageFrame(ImageFormat::Format format, int width, int height,
-             uint32 alignment_boundary);
+             uint32_t alignment_boundary);
   // Same as above, but use kDefaultAlignmentBoundary for alignment_boundary.
   ImageFrame(ImageFormat::Format format, int width, int height);
 
@@ -115,8 +115,8 @@ class ImageFrame {
   // width*num_channels*depth.  Both width_step and depth are in units
   // of bytes.
   ImageFrame(ImageFormat::Format format, int width, int height, int width_step,
-             uint8* pixel_data,
-             Deleter deleter = std::default_delete<uint8[]>());
+             uint8_t* pixel_data,
+             Deleter deleter = std::default_delete<uint8_t[]>());
 
   ImageFrame(ImageFrame&& move_from);
   ImageFrame& operator=(ImageFrame&& move_from);
@@ -142,7 +142,7 @@ class ImageFrame {
   // alignment_boundary.  If IsAligned(16) is true then so are
   // IsAligned(8), IsAligned(4), IsAligned(2), and IsAligned(1).
   // alignment_boundary must be 1 or a power of 2.
-  bool IsAligned(uint32 alignment_boundary) const;
+  bool IsAligned(uint32_t alignment_boundary) const;
 
   // Returns the image / video format.
   ImageFormat::Format Format() const { return format_; }
@@ -167,13 +167,13 @@ class ImageFrame {
   // Reset the current image frame and copy the data from image_frame into
   // this image frame.  The alignment_boundary must be given (and won't
   // necessarily match the alignment_boundary of the input image_frame).
-  void CopyFrom(const ImageFrame& image_frame, uint32 alignment_boundary);
+  void CopyFrom(const ImageFrame& image_frame, uint32_t alignment_boundary);
 
   // Get a mutable pointer to the underlying image data.  The ImageFrame
   // retains ownership.
-  uint8* MutablePixelData() { return pixel_data_.get(); }
+  uint8_t* MutablePixelData() { return pixel_data_.get(); }
   // Get a const pointer to the underlying image data.
-  const uint8* PixelData() const { return pixel_data_.get(); }
+  const uint8_t* PixelData() const { return pixel_data_.get(); }
 
   // Returns the total size of the pixel data.
   int PixelDataSize() const { return Height() * WidthStep(); }
@@ -187,41 +187,41 @@ class ImageFrame {
   // ImageFrame takes ownership of pixel_data.  See the Constructor
   // with the same arguments for details.
   void AdoptPixelData(ImageFormat::Format format, int width, int height,
-                      int width_step, uint8* pixel_data,
-                      Deleter deleter = std::default_delete<uint8[]>());
+                      int width_step, uint8_t* pixel_data,
+                      Deleter deleter = std::default_delete<uint8_t[]>());
 
   // Resets the ImageFrame and makes it a copy of the provided pixel
   // data, which is assumed to be stored contiguously.  The ImageFrame
   // will use the given alignment_boundary.
   void CopyPixelData(ImageFormat::Format format, int width, int height,
-                     const uint8* pixel_data, uint32 alignment_boundary);
+                     const uint8_t* pixel_data, uint32_t alignment_boundary);
 
   // Resets the ImageFrame and makes it a copy of the provided pixel
   // data, with given width_step.  The ImageFrame
   // will use the given alignment_boundary.
   void CopyPixelData(ImageFormat::Format format, int width, int height,
-                     int width_step, const uint8* pixel_data,
-                     uint32 alignment_boundary);
+                     int width_step, const uint8_t* pixel_data,
+                     uint32_t alignment_boundary);
 
   // Allocates a frame of the specified format, width, height, and alignment,
   // without clearing any current pixel data. See the constructor with the same
   // argument list.
   void Reset(ImageFormat::Format format, int width, int height,
-             uint32 alignment_boundary);
+             uint32_t alignment_boundary);
 
   // Relinquishes ownership of the pixel data.  Notice that the unique_ptr
   // uses a non-standard deleter.
-  std::unique_ptr<uint8[], Deleter> Release();
+  std::unique_ptr<uint8_t[], Deleter> Release();
 
   // Copy the 8-bit ImageFrame into a contiguous, pre-allocated buffer. Note
   // that ImageFrame does not necessarily store its data contiguously (i.e. do
   // not use copy_n to move image data).
-  void CopyToBuffer(uint8* buffer, int buffer_size) const;
+  void CopyToBuffer(uint8_t* buffer, int buffer_size) const;
 
   // A version of CopyToBuffer for 16-bit pixel data. Note that buffer_size
   // stores the number of 16-bit elements in the buffer, not the number of
   // bytes.
-  void CopyToBuffer(uint16* buffer, int buffer_size) const;
+  void CopyToBuffer(uint16_t* buffer, int buffer_size) const;
 
   // A version of CopyToBuffer for float pixel data. Note that buffer_size
   // stores the number of float elements in the buffer, not the number of
@@ -233,12 +233,12 @@ class ImageFrame {
 
  private:
   // Returns true if alignment_number is 1 or a power of 2.
-  static bool IsValidAlignmentNumber(uint32 alignment_boundary);
+  static bool IsValidAlignmentNumber(uint32_t alignment_boundary);
 
   // The internal implementation of copying data from the provided pixel data.
   // If width_step is 0, then calculates width_step assuming no padding.
   void InternalCopyFrom(int width, int height, int width_step, int channel_size,
-                        const uint8* pixel_data);
+                        const uint8_t* pixel_data);
 
   // The internal implementation of copying data to the provided buffer.
   // If width_step is 0, then calculates width_step assuming no padding.
@@ -249,7 +249,7 @@ class ImageFrame {
   int height_;
   int width_step_;
 
-  std::unique_ptr<uint8[], Deleter> pixel_data_;
+  std::unique_ptr<uint8_t[], Deleter> pixel_data_;
 };
 
 }  // namespace mediapipe

mediapipe/framework/formats/location.cc

@@ -16,6 +16,7 @@
 
 #include <algorithm>
 #include <cmath>
+#include <cstdint>
 #include <memory>
 
 #include "absl/log/absl_check.h"

mediapipe/framework/formats/location_opencv.cc

@@ -215,7 +215,7 @@ Location CreateCvMaskLocation(const cv::Mat_<T>& mask) {
   return Location(location_data);
 }
 
-template Location CreateCvMaskLocation(const cv::Mat_<uint8>& mask);
+template Location CreateCvMaskLocation(const cv::Mat_<uint8_t>& mask);
 template Location CreateCvMaskLocation(const cv::Mat_<float>& mask);
 
 }  // namespace mediapipe

mediapipe/framework/formats/tensor.cc

@@ -24,6 +24,9 @@
 #if MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
 #include "mediapipe/gpu/gl_base.h"
 #endif  // MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
+#ifdef MEDIAPIPE_TENSOR_USE_AHWB
+#include "mediapipe/framework/formats/hardware_buffer.h"
+#endif  // MEDIAPIPE_TENSOR_USE_AHWB
 
 #if MEDIAPIPE_METAL_ENABLED
 #import <Metal/Metal.h>
@@ -339,6 +342,14 @@ Tensor::OpenGlBufferView Tensor::GetOpenGlBufferReadView() const {
       << "Tensor conversion between different GPU backing formats is not "
          "supported yet.";
   auto lock(absl::make_unique<absl::MutexLock>(&view_mutex_));
+  if ((valid_ & kValidOpenGlBuffer) && gl_context_ != nullptr &&
+      !gl_context_->IsCurrent() && GlContext::IsAnyContextCurrent()) {
+    ABSL_LOG_FIRST_N(WARNING, 1)
+        << "Tensor::GetOpenGlBufferReadView is not executed on the same GL "
+           "context where GL buffer was created. Note that Tensor has "
+           "limited synchronization support when sharing OpenGl objects "
+           "between multiple OpenGL contexts.";
+  }
   AllocateOpenGlBuffer();
   if (!(valid_ & kValidOpenGlBuffer)) {
     // If the call succeeds then AHWB -> SSBO are synchronized so any usage of
@@ -356,7 +367,13 @@ Tensor::OpenGlBufferView Tensor::GetOpenGlBufferReadView() const {
   }
   return {opengl_buffer_, std::move(lock),
 #ifdef MEDIAPIPE_TENSOR_USE_AHWB
-          &ssbo_read_
+          // ssbo_read_ is passed to be populated on OpenGlBufferView
+          // destruction in order to perform delayed resources releasing (see
+          // tensor_ahwb.cc/DelayedReleaser) only when AHWB is in use.
+          //
+          // Not passing for the case when AHWB is not in use to avoid creation
+          // of unnecessary sync object and memory leak.
+          use_ahwb_ ? &ssbo_read_ : nullptr
 #else
           nullptr
 #endif  // MEDIAPIPE_TENSOR_USE_AHWB
@@ -367,6 +384,14 @@ Tensor::OpenGlBufferView Tensor::GetOpenGlBufferWriteView(
     uint64_t source_location_hash) const {
   auto lock(absl::make_unique<absl::MutexLock>(&view_mutex_));
   TrackAhwbUsage(source_location_hash);
+  if ((valid_ & kValidOpenGlBuffer) && gl_context_ != nullptr &&
+      !gl_context_->IsCurrent() && GlContext::IsAnyContextCurrent()) {
+    ABSL_LOG_FIRST_N(WARNING, 1)
+        << "Tensor::GetOpenGlBufferWriteView is not executed on the same GL "
+           "context where GL buffer was created. Note that Tensor has "
+           "limited synchronization support when sharing OpenGl objects "
+           "between multiple OpenGL contexts.";
+  }
   AllocateOpenGlBuffer();
   valid_ = kValidOpenGlBuffer;
   return {opengl_buffer_, std::move(lock), nullptr};
@@ -536,9 +561,8 @@ Tensor::CpuReadView Tensor::GetCpuReadView() const {
     void* ptr = MapAhwbToCpuRead();
     if (ptr) {
       valid_ |= kValidCpu;
-      return {ptr, std::move(lock), [ahwb = ahwb_] {
-                auto error = AHardwareBuffer_unlock(ahwb, nullptr);
-                ABSL_CHECK(error == 0) << "AHardwareBuffer_unlock " << error;
+      return {ptr, std::move(lock), [ahwb = ahwb_.get()] {
+                ABSL_CHECK_OK(ahwb->Unlock()) << "Unlock failed.";
               }};
     }
   }
@@ -620,9 +644,11 @@ Tensor::CpuWriteView Tensor::GetCpuWriteView(
   if (__builtin_available(android 26, *)) {
     void* ptr = MapAhwbToCpuWrite();
     if (ptr) {
-      return {ptr, std::move(lock), [ahwb = ahwb_, fence_fd = &fence_fd_] {
-                auto error = AHardwareBuffer_unlock(ahwb, fence_fd);
-                ABSL_CHECK(error == 0) << "AHardwareBuffer_unlock " << error;
+      return {ptr, std::move(lock),
+              [ahwb = ahwb_.get(), fence_fd = &fence_fd_] {
+                auto fence_fd_status = ahwb->UnlockAsync();
+                ABSL_CHECK_OK(fence_fd_status) << "Unlock failed.";
+                *fence_fd = fence_fd_status.value();
               }};
     }
   }

mediapipe/framework/formats/tensor.h

@@ -44,7 +44,8 @@
 #ifdef MEDIAPIPE_TENSOR_USE_AHWB
 #include <EGL/egl.h>
 #include <EGL/eglext.h>
-#include <android/hardware_buffer.h>
+
+#include "mediapipe/framework/formats/hardware_buffer.h"
 #endif  // MEDIAPIPE_TENSOR_USE_AHWB
 #if MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
 #include "mediapipe/gpu/gl_base.h"
@@ -195,9 +196,11 @@ class Tensor {
   using FinishingFunc = std::function<bool(bool)>;
   class AHardwareBufferView : public View {
    public:
-    AHardwareBuffer* handle() const { return handle_; }
+    AHardwareBuffer* handle() const {
+      return hardware_buffer_->GetAHardwareBuffer();
+    }
     AHardwareBufferView(AHardwareBufferView&& src) : View(std::move(src)) {
-      handle_ = std::exchange(src.handle_, nullptr);
+      hardware_buffer_ = std::move(src.hardware_buffer_);
       file_descriptor_ = src.file_descriptor_;
       fence_fd_ = std::exchange(src.fence_fd_, nullptr);
       ahwb_written_ = std::exchange(src.ahwb_written_, nullptr);
@@ -222,17 +225,17 @@ class Tensor {
 
    protected:
     friend class Tensor;
-    AHardwareBufferView(AHardwareBuffer* handle, int file_descriptor,
+    AHardwareBufferView(HardwareBuffer* hardware_buffer, int file_descriptor,
                         int* fence_fd, FinishingFunc* ahwb_written,
                         std::function<void()>* release_callback,
                         std::unique_ptr<absl::MutexLock>&& lock)
         : View(std::move(lock)),
-          handle_(handle),
+          hardware_buffer_(hardware_buffer),
           file_descriptor_(file_descriptor),
           fence_fd_(fence_fd),
           ahwb_written_(ahwb_written),
           release_callback_(release_callback) {}
-    AHardwareBuffer* handle_;
+    HardwareBuffer* hardware_buffer_;
     int file_descriptor_;
     // The view sets some Tensor's fields. The view is released prior to tensor.
     int* fence_fd_;
@@ -285,18 +288,22 @@ class Tensor {
   class OpenGlBufferView : public View {
    public:
     GLuint name() const { return name_; }
+
     OpenGlBufferView(OpenGlBufferView&& src) : View(std::move(src)) {
       name_ = std::exchange(src.name_, GL_INVALID_INDEX);
       ssbo_read_ = std::exchange(src.ssbo_read_, nullptr);
     }
     ~OpenGlBufferView() {
       if (ssbo_read_) {
+        // TODO: update tensor to properly handle cases when
+        // multiple views were requested multiple sync fence may be needed.
         *ssbo_read_ = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
       }
     }
 
    protected:
     friend class Tensor;
+
     OpenGlBufferView(GLuint name, std::unique_ptr<absl::MutexLock>&& lock,
                      GLsync* ssbo_read)
         : View(std::move(lock)), name_(name), ssbo_read_(ssbo_read) {}
@@ -384,7 +391,7 @@ class Tensor {
   mutable std::unique_ptr<MtlResources> mtl_resources_;
 
 #ifdef MEDIAPIPE_TENSOR_USE_AHWB
-  mutable AHardwareBuffer* ahwb_ = nullptr;
+  mutable std::unique_ptr<HardwareBuffer> ahwb_;
   // Signals when GPU finished writing into SSBO so AHWB can be used then. Or
   // signals when writing into AHWB has been finished so GPU can read from SSBO.
   // Sync and FD are bound together.

mediapipe/framework/formats/tensor_ahwb.cc

@@ -10,7 +10,7 @@
 #include "absl/log/absl_check.h"
 #include "absl/log/absl_log.h"
 #include "absl/synchronization/mutex.h"
-#include "mediapipe/framework/port.h"
+#include "mediapipe/framework/formats/hardware_buffer.h"
 #include "mediapipe/gpu/gl_base.h"
 #endif  // MEDIAPIPE_TENSOR_USE_AHWB
 
@@ -97,7 +97,7 @@ class DelayedReleaser {
   DelayedReleaser(DelayedReleaser&&) = delete;
   DelayedReleaser& operator=(DelayedReleaser&&) = delete;
 
-  static void Add(AHardwareBuffer* ahwb, GLuint opengl_buffer,
+  static void Add(std::unique_ptr<HardwareBuffer> ahwb, GLuint opengl_buffer,
                   EGLSyncKHR ssbo_sync, GLsync ssbo_read,
                   Tensor::FinishingFunc&& ahwb_written,
                   std::shared_ptr<mediapipe::GlContext> gl_context,
@@ -115,8 +115,8 @@ class DelayedReleaser {
 
     // Using `new` to access a non-public constructor.
     to_release_local.emplace_back(absl::WrapUnique(new DelayedReleaser(
-        ahwb, opengl_buffer, ssbo_sync, ssbo_read, std::move(ahwb_written),
-        gl_context, std::move(callback))));
+        std::move(ahwb), opengl_buffer, ssbo_sync, ssbo_read,
+        std::move(ahwb_written), gl_context, std::move(callback))));
     for (auto it = to_release_local.begin(); it != to_release_local.end();) {
       if ((*it)->IsSignaled()) {
         it = to_release_local.erase(it);
@@ -136,9 +136,6 @@ class DelayedReleaser {
 
   ~DelayedReleaser() {
     if (release_callback_) release_callback_();
-    if (__builtin_available(android 26, *)) {
-      AHardwareBuffer_release(ahwb_);
-    }
   }
 
   bool IsSignaled() {
@@ -181,7 +178,7 @@ class DelayedReleaser {
   }
 
  protected:
-  AHardwareBuffer* ahwb_;
+  std::unique_ptr<HardwareBuffer> ahwb_;
   GLuint opengl_buffer_;
   // TODO: use wrapper instead.
   EGLSyncKHR fence_sync_;
@@ -192,12 +189,12 @@ class DelayedReleaser {
   std::function<void()> release_callback_;
   static inline std::deque<std::unique_ptr<DelayedReleaser>> to_release_;
 
-  DelayedReleaser(AHardwareBuffer* ahwb, GLuint opengl_buffer,
+  DelayedReleaser(std::unique_ptr<HardwareBuffer> ahwb, GLuint opengl_buffer,
                   EGLSyncKHR fence_sync, GLsync ssbo_read,
                   Tensor::FinishingFunc&& ahwb_written,
                   std::shared_ptr<mediapipe::GlContext> gl_context,
                   std::function<void()>&& callback)
-      : ahwb_(ahwb),
+      : ahwb_(std::move(ahwb)),
         opengl_buffer_(opengl_buffer),
         fence_sync_(fence_sync),
         ssbo_read_(ssbo_read),
@@ -214,7 +211,7 @@ Tensor::AHardwareBufferView Tensor::GetAHardwareBufferReadView() const {
   ABSL_CHECK(!(valid_ & kValidOpenGlTexture2d))
       << "Tensor conversion between OpenGL texture and AHardwareBuffer is not "
          "supported.";
-  bool transfer = !ahwb_;
+  bool transfer = ahwb_ == nullptr;
   ABSL_CHECK(AllocateAHardwareBuffer())
       << "AHardwareBuffer is not supported on the target system.";
   valid_ |= kValidAHardwareBuffer;
@@ -223,12 +220,10 @@ Tensor::AHardwareBufferView Tensor::GetAHardwareBufferReadView() const {
   } else {
     if (valid_ & kValidOpenGlBuffer) CreateEglSyncAndFd();
   }
-  return {ahwb_,
-          ssbo_written_,
+  return {ahwb_.get(),        ssbo_written_,
           &fence_fd_,  // The FD is created for SSBO -> AHWB synchronization.
           &ahwb_written_,  // Filled by SetReadingFinishedFunc.
-          &release_callback_,
-          std::move(lock)};
+          &release_callback_, std::move(lock)};
 }
 
 void Tensor::CreateEglSyncAndFd() const {
@@ -258,12 +253,11 @@ Tensor::AHardwareBufferView Tensor::GetAHardwareBufferWriteView(
   ABSL_CHECK(AllocateAHardwareBuffer(size_alignment))
       << "AHardwareBuffer is not supported on the target system.";
   valid_ = kValidAHardwareBuffer;
-  return {ahwb_,
+  return {ahwb_.get(),
           /*ssbo_written=*/-1,
           &fence_fd_,      // For SetWritingFinishedFD.
-          &ahwb_written_,
-          &release_callback_,
-          std::move(lock)};
+          &ahwb_written_,  // Filled by SetReadingFinishedFunc.
+          &release_callback_,  std::move(lock)};
 }
 
 bool Tensor::AllocateAHardwareBuffer(int size_alignment) const {
@@ -276,40 +270,43 @@ bool Tensor::AllocateAHardwareBuffer(int size_alignment) const {
   }
   use_ahwb_ = true;
 
-  if (__builtin_available(android 26, *)) {
   if (ahwb_ == nullptr) {
-      AHardwareBuffer_Desc desc = {};
+    HardwareBufferSpec spec = {};
     if (size_alignment == 0) {
-        desc.width = bytes();
+      spec.width = bytes();
     } else {
       // We expect allocations to be page-aligned, implicitly satisfying any
       // requirements from Edge TPU. No need to add a check for this,
       // since Edge TPU will check for us.
-        desc.width = AlignedToPowerOf2(bytes(), size_alignment);
+      spec.width = AlignedToPowerOf2(bytes(), size_alignment);
     }
-      desc.height = 1;
-      desc.layers = 1;
-      desc.format = AHARDWAREBUFFER_FORMAT_BLOB;
-      desc.usage = AHARDWAREBUFFER_USAGE_CPU_WRITE_OFTEN |
-                   AHARDWAREBUFFER_USAGE_CPU_READ_OFTEN |
-                   AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER;
-      return AHardwareBuffer_allocate(&desc, &ahwb_) == 0;
+    spec.height = 1;
+    spec.layers = 1;
+    spec.format = HardwareBufferSpec::AHARDWAREBUFFER_FORMAT_BLOB;
+    spec.usage = HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_WRITE_OFTEN |
+                 HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_READ_OFTEN |
+                 HardwareBufferSpec::AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER;
+    auto new_ahwb = HardwareBuffer::Create(spec);
+    if (!new_ahwb.ok()) {
+      ABSL_LOG(ERROR) << "Allocation of NDK Hardware Buffer failed: "
+                      << new_ahwb.status();
+      return false;
+    }
+    ahwb_ = std::make_unique<HardwareBuffer>(std::move(*new_ahwb));
   }
   return true;
 }
-  return false;
-}
 
 bool Tensor::AllocateAhwbMapToSsbo() const {
   if (__builtin_available(android 26, *)) {
     if (AllocateAHardwareBuffer()) {
-      if (MapAHardwareBufferToGlBuffer(ahwb_, bytes()).ok()) {
+      if (MapAHardwareBufferToGlBuffer(ahwb_->GetAHardwareBuffer(), bytes())
+              .ok()) {
         glBindBuffer(GL_SHADER_STORAGE_BUFFER, 0);
         return true;
       }
       // Unable to make OpenGL <-> AHWB binding. Use regular SSBO instead.
-      AHardwareBuffer_release(ahwb_);
-      ahwb_ = nullptr;
+      ahwb_.reset();
     }
   }
   return false;
@@ -317,14 +314,11 @@ bool Tensor::AllocateAhwbMapToSsbo() const {
 
 // Moves Cpu/Ssbo resource under the Ahwb backed memory.
 void Tensor::MoveCpuOrSsboToAhwb() const {
-  void* dest = nullptr;
-  if (__builtin_available(android 26, *)) {
-    auto error = AHardwareBuffer_lock(
-        ahwb_, AHARDWAREBUFFER_USAGE_CPU_WRITE_RARELY, -1, nullptr, &dest);
-    ABSL_CHECK(error == 0) << "AHardwareBuffer_lock " << error;
-  }
+  auto dest =
+      ahwb_->Lock(HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_WRITE_RARELY);
+  ABSL_CHECK_OK(dest) << "Lock of AHWB failed";
   if (valid_ & kValidCpu) {
-    std::memcpy(dest, cpu_buffer_, bytes());
+    std::memcpy(*dest, cpu_buffer_, bytes());
     // Free CPU memory because next time AHWB is mapped instead.
     free(cpu_buffer_);
     cpu_buffer_ = nullptr;
@@ -334,7 +328,7 @@ void Tensor::MoveCpuOrSsboToAhwb() const {
       glBindBuffer(GL_SHADER_STORAGE_BUFFER, opengl_buffer_);
       const void* src = glMapBufferRange(GL_SHADER_STORAGE_BUFFER, 0, bytes(),
                                          GL_MAP_READ_BIT);
-      std::memcpy(dest, src, bytes());
+      std::memcpy(*dest, src, bytes());
       glUnmapBuffer(GL_SHADER_STORAGE_BUFFER);
       glDeleteBuffers(1, &opengl_buffer_);
     });
@@ -347,10 +341,7 @@ void Tensor::MoveCpuOrSsboToAhwb() const {
     ABSL_LOG(FATAL) << "Can't convert tensor with mask " << valid_
                     << " into AHWB.";
   }
-  if (__builtin_available(android 26, *)) {
-    auto error = AHardwareBuffer_unlock(ahwb_, nullptr);
-    ABSL_CHECK(error == 0) << "AHardwareBuffer_unlock " << error;
-  }
+  ABSL_CHECK_OK(ahwb_->Unlock()) << "Unlock of AHWB failed";
 }
 
 // SSBO is created on top of AHWB. A fence is inserted into the GPU queue before
@@ -403,59 +394,52 @@ void Tensor::ReleaseAhwbStuff() {
     if (ahwb_) {
       if (ssbo_read_ != 0 || fence_sync_ != EGL_NO_SYNC_KHR || ahwb_written_) {
         if (ssbo_written_ != -1) close(ssbo_written_);
-        DelayedReleaser::Add(ahwb_, opengl_buffer_, fence_sync_, ssbo_read_,
-                             std::move(ahwb_written_), gl_context_,
+        DelayedReleaser::Add(std::move(ahwb_), opengl_buffer_, fence_sync_,
+                             ssbo_read_, std::move(ahwb_written_), gl_context_,
                              std::move(release_callback_));
         opengl_buffer_ = GL_INVALID_INDEX;
       } else {
         if (release_callback_) release_callback_();
-        AHardwareBuffer_release(ahwb_);
+        ahwb_.reset();
       }
     }
   }
 }
 
 void* Tensor::MapAhwbToCpuRead() const {
-  if (__builtin_available(android 26, *)) {
-    if (ahwb_) {
+  if (ahwb_ != nullptr) {
     if (!(valid_ & kValidCpu)) {
       if ((valid_ & kValidOpenGlBuffer) && ssbo_written_ == -1) {
         // EGLSync is failed. Use another synchronization method.
         // TODO: Use tflite::gpu::GlBufferSync and GlActiveSync.
         gl_context_->Run([]() { glFinish(); });
       } else if (valid_ & kValidAHardwareBuffer) {
-          ABSL_CHECK(ahwb_written_)
-              << "Ahwb-to-Cpu synchronization requires the "
+        ABSL_CHECK(ahwb_written_) << "Ahwb-to-Cpu synchronization requires the "
                                      "completion function to be set";
         ABSL_CHECK(ahwb_written_(true))
             << "An error oqcured while waiting for the buffer to be written";
       }
     }
-      void* ptr;
-      auto error =
-          AHardwareBuffer_lock(ahwb_, AHARDWAREBUFFER_USAGE_CPU_READ_OFTEN,
-                               ssbo_written_, nullptr, &ptr);
-      ABSL_CHECK(error == 0) << "AHardwareBuffer_lock " << error;
+    auto ptr =
+        ahwb_->Lock(HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_READ_OFTEN,
+                    ssbo_written_);
+    ABSL_CHECK_OK(ptr) << "Lock of AHWB failed";
     close(ssbo_written_);
     ssbo_written_ = -1;
-      return ptr;
-    }
+    return *ptr;
   }
   return nullptr;
 }
 
 void* Tensor::MapAhwbToCpuWrite() const {
-  if (__builtin_available(android 26, *)) {
-    if (ahwb_) {
+  if (ahwb_ != nullptr) {
     // TODO: If previously acquired view is GPU write view then need
     // to be sure that writing is finished. That's a warning: two consequent
     // write views should be interleaved with read view.
-      void* ptr;
-      auto error = AHardwareBuffer_lock(
-          ahwb_, AHARDWAREBUFFER_USAGE_CPU_WRITE_OFTEN, -1, nullptr, &ptr);
-      ABSL_CHECK(error == 0) << "AHardwareBuffer_lock " << error;
-      return ptr;
-    }
+    auto locked_ptr =
+        ahwb_->Lock(HardwareBufferSpec::AHARDWAREBUFFER_USAGE_CPU_WRITE_OFTEN);
+    ABSL_CHECK_OK(locked_ptr) << "Lock of AHWB failed";
+    return *locked_ptr;
   }
   return nullptr;
 }

mediapipe/framework/formats/tensor_ahwb_gpu_test.cc

@@ -6,6 +6,7 @@
 
 #include <cstdint>
 
+#include "absl/algorithm/container.h"
 #include "mediapipe/framework/formats/tensor.h"
 #include "mediapipe/framework/formats/tensor/views/data_types.h"
 #include "mediapipe/gpu/gpu_test_base.h"
@@ -18,7 +19,7 @@
 // Then the test requests the CPU view and compares the values.
 // Float32 and Float16 tests are there.
 
-namespace {
+namespace mediapipe {
 
 using mediapipe::Float16;
 using mediapipe::Tensor;
@@ -27,6 +28,16 @@ MATCHER_P(NearWithPrecision, precision, "") {
   return std::abs(std::get<0>(arg) - std::get<1>(arg)) < precision;
 }
 
+template <typename F = float>
+std::vector<F> CreateReferenceData(int num_elements) {
+  std::vector<F> reference;
+  reference.resize(num_elements);
+  for (int i = 0; i < num_elements; i++) {
+    reference[i] = static_cast<float>(i) / 10.0f;
+  }
+  return reference;
+}
+
 #if MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_31
 
 // Utility function to fill the GPU buffer.
@@ -110,11 +121,7 @@ TEST_F(TensorAhwbGpuTest, TestGpuToCpuFloat32) {
   });
   auto ptr = tensor.GetCpuReadView().buffer<float>();
   ASSERT_NE(ptr, nullptr);
-  std::vector<float> reference;
-  reference.resize(num_elements);
-  for (int i = 0; i < num_elements; i++) {
-    reference[i] = static_cast<float>(i) / 10.0f;
-  }
+  std::vector<float> reference = CreateReferenceData(num_elements);
   EXPECT_THAT(absl::Span<const float>(ptr, num_elements),
               testing::Pointwise(testing::FloatEq(), reference));
 }
@@ -137,11 +144,7 @@ TEST_F(TensorAhwbGpuTest, TestGpuToCpuFloat16) {
   });
   auto ptr = tensor.GetCpuReadView().buffer<Float16>();
   ASSERT_NE(ptr, nullptr);
-  std::vector<Float16> reference;
-  reference.resize(num_elements);
-  for (int i = 0; i < num_elements; i++) {
-    reference[i] = static_cast<float>(i) / 10.0f;
-  }
+  std::vector<Float16> reference = CreateReferenceData<Float16>(num_elements);
   // Precision is set to a reasonable value for Float16.
   EXPECT_THAT(absl::Span<const Float16>(ptr, num_elements),
               testing::Pointwise(NearWithPrecision(0.001), reference));
@@ -166,11 +169,7 @@ TEST_F(TensorAhwbGpuTest, TestReplacingCpuByAhwb) {
   }
   auto ptr = tensor.GetCpuReadView().buffer<float>();
   ASSERT_NE(ptr, nullptr);
-  std::vector<float> reference;
-  reference.resize(num_elements);
-  for (int i = 0; i < num_elements; i++) {
-    reference[i] = static_cast<float>(i) / 10.0f;
-  }
+  std::vector<float> reference = CreateReferenceData(num_elements);
   EXPECT_THAT(absl::Span<const float>(ptr, num_elements),
               testing::Pointwise(testing::FloatEq(), reference));
 }
@@ -194,17 +193,107 @@ TEST_F(TensorAhwbGpuTest, TestReplacingGpuByAhwb) {
   }
   auto ptr = tensor.GetCpuReadView().buffer<float>();
   ASSERT_NE(ptr, nullptr);
-  std::vector<float> reference;
-  reference.resize(num_elements);
-  for (int i = 0; i < num_elements; i++) {
-    reference[i] = static_cast<float>(i) / 10.0f;
-  }
+  std::vector<float> reference = CreateReferenceData(num_elements);
   EXPECT_THAT(absl::Span<const float>(ptr, num_elements),
               testing::Pointwise(testing::FloatEq(), reference));
 }
 
+std::vector<float> ReadGlBufferView(const Tensor::OpenGlBufferView& view,
+                                    int num_elements) {
+  glBindBuffer(GL_SHADER_STORAGE_BUFFER, view.name());
+  int bytes = num_elements * sizeof(float);
+  void* ptr =
+      glMapBufferRange(GL_SHADER_STORAGE_BUFFER, 0, bytes, GL_MAP_READ_BIT);
+  ABSL_CHECK(ptr) << "glMapBufferRange failed: " << glGetError();
+
+  std::vector<float> data;
+  data.resize(num_elements);
+  std::memcpy(data.data(), ptr, bytes);
+  glUnmapBuffer(GL_SHADER_STORAGE_BUFFER);
+  return data;
+}
+
+TEST_F(TensorAhwbGpuTest, TestGetOpenGlBufferReadViewNoAhwb) {
+  constexpr size_t kNumElements = 20;
+  std::vector<float> reference = CreateReferenceData(kNumElements);
+
+  Tensor tensor(Tensor::ElementType::kFloat32, Tensor::Shape({kNumElements}));
+  {
+    // Populate tensor on CPU and make sure view is destroyed
+    absl::c_copy(reference, tensor.GetCpuWriteView().buffer<float>());
+  }
+
+  RunInGlContext([&] {
+    // Triggers conversion to GL buffer.
+    auto ssbo_view = tensor.GetOpenGlBufferReadView();
+    ASSERT_NE(ssbo_view.name(), 0);
+    // ssbo_read_ must NOT be populated, as there's no AHWB associated with
+    // GL buffer
+    ASSERT_EQ(ssbo_view.ssbo_read_, nullptr);
+
+    std::vector<float> output = ReadGlBufferView(ssbo_view, kNumElements);
+    EXPECT_THAT(output, testing::Pointwise(testing::FloatEq(), reference));
+  });
+}
+
+TEST_F(TensorAhwbGpuTest, TestGetOpenGlBufferReadViewAhwbFromCpu) {
+  constexpr size_t kNumElements = 20;
+  std::vector<float> reference = CreateReferenceData(kNumElements);
+
+  Tensor tensor(Tensor::ElementType::kFloat32, Tensor::Shape({kNumElements}));
+  {
+    // Populate tensor on CPU and make sure view is destroyed
+    absl::c_copy(reference, tensor.GetCpuWriteView().buffer<float>());
+  }
+
+  {
+    // Make tensor to allocate ahwb and make sure view is destroyed.
+    ASSERT_NE(tensor.GetAHardwareBufferReadView().handle(), nullptr);
+  }
+
+  RunInGlContext([&] {
+    // Triggers conversion to GL buffer.
+    auto ssbo_view = tensor.GetOpenGlBufferReadView();
+    ASSERT_NE(ssbo_view.name(), 0);
+    // ssbo_read_ must be populated, so during view destruction it's set
+    // properly for further AHWB destruction
+    ASSERT_NE(ssbo_view.ssbo_read_, nullptr);
+
+    std::vector<float> output = ReadGlBufferView(ssbo_view, kNumElements);
+    EXPECT_THAT(output, testing::Pointwise(testing::FloatEq(), reference));
+  });
+}
+
+TEST_F(TensorAhwbGpuTest, TestGetOpenGlBufferReadViewAhwbFromGpu) {
+  constexpr size_t kNumElements = 20;
+  std::vector<float> reference = CreateReferenceData(kNumElements);
+
+  Tensor tensor(Tensor::ElementType::kFloat32, Tensor::Shape({kNumElements}));
+  {
+    // Make tensor to allocate ahwb and make sure view is destroyed.
+    ASSERT_NE(tensor.GetAHardwareBufferWriteView().handle(), nullptr);
+  }
+
+  RunInGlContext([&] {
+    FillGpuBuffer(tensor.GetOpenGlBufferWriteView().name(),
+                  tensor.shape().num_elements(), tensor.element_type());
+  });
+
+  RunInGlContext([&] {
+    // Triggers conversion to GL buffer.
+    auto ssbo_view = tensor.GetOpenGlBufferReadView();
+    ASSERT_NE(ssbo_view.name(), 0);
+    // ssbo_read_ must be populated, so during view destruction it's set
+    // properly for further AHWB destruction
+    ASSERT_NE(ssbo_view.ssbo_read_, nullptr);
+
+    std::vector<float> output = ReadGlBufferView(ssbo_view, kNumElements);
+    EXPECT_THAT(output, testing::Pointwise(testing::FloatEq(), reference));
+  });
+}
+
 #endif  // MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_31
-}  // namespace
+}  // namespace mediapipe
 
 #endif  // !defined(MEDIAPIPE_NO_JNI) && (__ANDROID_API__ >= 26 ||
         // defined(__ANDROID_UNAVAILABLE_SYMBOLS_ARE_WEAK__))

mediapipe/framework/formats/tensor_ahwb_test.cc

@@ -1,3 +1,5 @@
+#include <android/hardware_buffer.h>
+
 #include "mediapipe/framework/formats/tensor.h"
 #include "testing/base/public/gmock.h"
 #include "testing/base/public/gunit.h"

mediapipe/framework/formats/yuv_image.h

@@ -15,6 +15,7 @@
 #ifndef MEDIAPIPE_FRAMEWORK_FORMATS_YUV_IMAGE_H_
 #define MEDIAPIPE_FRAMEWORK_FORMATS_YUV_IMAGE_H_
 
+#include <cstdint>
 #include <functional>
 #include <memory>
 
@@ -120,12 +121,12 @@ class YUVImage {
 
   // Convenience constructor
   YUVImage(libyuv::FourCC fourcc,                     //
-           std::unique_ptr<uint8[]> data_location,  //
-           uint8* data0, int stride0,               //
-           uint8* data1, int stride1,               //
-           uint8* data2, int stride2,               //
+           std::unique_ptr<uint8_t[]> data_location,  //
+           uint8_t* data0, int stride0,               //
+           uint8_t* data1, int stride1,               //
+           uint8_t* data2, int stride2,               //
            int width, int height, int bit_depth = 8) {
-    uint8* tmp = data_location.release();
+    uint8_t* tmp = data_location.release();
     std::function<void()> deallocate = [tmp]() { delete[] tmp; };
     Initialize(fourcc,          //
                deallocate,      //
@@ -138,13 +139,13 @@ class YUVImage {
   // Convenience constructor to construct the YUVImage with data stored
   // in three unique_ptrs.
   YUVImage(libyuv::FourCC fourcc,                          //
-           std::unique_ptr<uint8[]> data0, int stride0,  //
-           std::unique_ptr<uint8[]> data1, int stride1,  //
-           std::unique_ptr<uint8[]> data2, int stride2,  //
+           std::unique_ptr<uint8_t[]> data0, int stride0,  //
+           std::unique_ptr<uint8_t[]> data1, int stride1,  //
+           std::unique_ptr<uint8_t[]> data2, int stride2,  //
            int width, int height, int bit_depth = 8) {
-    uint8* tmp0 = data0.release();
-    uint8* tmp1 = data1.release();
-    uint8* tmp2 = data2.release();
+    uint8_t* tmp0 = data0.release();
+    uint8_t* tmp1 = data1.release();
+    uint8_t* tmp2 = data2.release();
     std::function<void()> deallocate = [tmp0, tmp1, tmp2]() {
       delete[] tmp0;
       delete[] tmp1;
@@ -177,9 +178,9 @@ class YUVImage {
   // pixel format it holds.
   void Initialize(libyuv::FourCC fourcc,                        //
                   std::function<void()> deallocation_function,  //
-                  uint8* data0, int stride0,                    //
-                  uint8* data1, int stride1,                    //
-                  uint8* data2, int stride2,                    //
+                  uint8_t* data0, int stride0,                  //
+                  uint8_t* data1, int stride1,                  //
+                  uint8_t* data2, int stride2,                  //
                   int width, int height, int bit_depth = 8) {
     Clear();
     deallocation_function_ = deallocation_function;
@@ -214,7 +215,7 @@ class YUVImage {
 
   // Getters.
   libyuv::FourCC fourcc() const { return fourcc_; }
-  const uint8* data(int index) const { return data_[index]; }
+  const uint8_t* data(int index) const { return data_[index]; }
   int stride(int index) const { return stride_[index]; }
   int width() const { return width_; }
   int height() const { return height_; }
@@ -226,7 +227,7 @@ class YUVImage {
 
   // Setters.
   void set_fourcc(libyuv::FourCC fourcc) { fourcc_ = fourcc; }
-  uint8* mutable_data(int index) { return data_[index]; }
+  uint8_t* mutable_data(int index) { return data_[index]; }
   void set_stride(int index, int stride) { stride_[index] = stride; }
   void set_width(int width) { width_ = width; }
   void set_height(int height) { height_ = height; }
@@ -241,7 +242,7 @@ class YUVImage {
   std::function<void()> deallocation_function_;
 
   libyuv::FourCC fourcc_ = libyuv::FOURCC_ANY;
-  uint8* data_[kMaxNumPlanes];
+  uint8_t* data_[kMaxNumPlanes];
   int stride_[kMaxNumPlanes];
   int width_ = 0;
   int height_ = 0;

mediapipe/framework/port/build_config.bzl

@@ -26,7 +26,7 @@ def replace_suffix(string, old, new):
 
 def mediapipe_ts_library(
         name,
-        srcs,
+        srcs = [],
         visibility = None,
         deps = [],
         testonly = 0,

mediapipe/gpu/BUILD

@@ -196,6 +196,7 @@ cc_library(
         ":gpu_buffer_format",
         "//mediapipe/framework:executor",
         "//mediapipe/framework:mediapipe_profiling",
+        "//mediapipe/framework:port",
         "//mediapipe/framework:timestamp",
         "//mediapipe/framework/port:logging",
         "//mediapipe/framework/port:ret_check",
@@ -209,6 +210,7 @@ cc_library(
         "@com_google_absl//absl/memory",
         "@com_google_absl//absl/status",
         "@com_google_absl//absl/status:statusor",
+        "@com_google_absl//absl/strings:str_format",
         "@com_google_absl//absl/synchronization",
     ] + select({
         "//conditions:default": [],

mediapipe/gpu/gl_context.cc

@@ -26,7 +26,9 @@
 #include "absl/log/absl_log.h"
 #include "absl/memory/memory.h"
 #include "absl/status/status.h"
+#include "absl/strings/str_format.h"
 #include "absl/synchronization/mutex.h"
+#include "mediapipe/framework/port.h"  // IWYU pragma: keep
 #include "mediapipe/framework/port/ret_check.h"
 #include "mediapipe/framework/port/status.h"
 #include "mediapipe/framework/port/status_builder.h"
@@ -48,6 +50,17 @@
 
 namespace mediapipe {
 
+namespace internal_gl_context {
+
+bool IsOpenGlVersionSameOrAbove(const OpenGlVersion& version,
+                                const OpenGlVersion& expected_version) {
+  return (version.major == expected_version.major &&
+          version.minor >= expected_version.minor) ||
+         version.major > expected_version.major;
+}
+
+}  // namespace internal_gl_context
+
 static void SetThreadName(const char* name) {
 #if defined(__GLIBC_PREREQ)
 #define LINUX_STYLE_SETNAME_NP __GLIBC_PREREQ(2, 12)
@@ -638,6 +651,11 @@ class GlSyncWrapper {
     // TODO: do something if the wait fails?
   }
 
+  // This method exists only for investigation purposes to distinguish stack
+  // traces: external vs. internal context.
+  // TODO: remove after glWaitSync crashes are resolved.
+  void WaitOnGpuExternalContext() { glWaitSync(sync_, 0, GL_TIMEOUT_IGNORED); }
+
   void WaitOnGpu() {
     if (!sync_) return;
       // WebGL2 specifies a waitSync call, but since cross-context
@@ -645,6 +663,33 @@ class GlSyncWrapper {
       // a warning when it's called, so let's just skip the call. See
       // b/184637485 for details.
 #ifndef __EMSCRIPTEN__
+
+    if (!GlContext::IsAnyContextCurrent()) {
+      // glWaitSync must be called on with some context current. Doing the
+      // opposite doesn't necessarily result in a crash or GL error. Hence,
+      // just logging an error and skipping the call.
+      ABSL_LOG_FIRST_N(ERROR, 1)
+          << "An attempt to wait for a sync without any context current.";
+      return;
+    }
+
+    auto context = GlContext::GetCurrent();
+    if (context == nullptr) {
+      // This can happen when WaitOnGpu is invoked on an external context,
+      // created by other than GlContext::Create means.
+      WaitOnGpuExternalContext();
+      return;
+    }
+
+    // GlContext::ShouldUseFenceSync guards creation of sync objects, so this
+    // CHECK should never fail if clients use MediaPipe APIs in an intended way.
+    // TODO: remove after glWaitSync crashes are resolved.
+    ABSL_CHECK(context->ShouldUseFenceSync()) << absl::StrFormat(
+        "An attempt to wait for a sync when it should not be used. (OpenGL "
+        "Version "
+        "%d.%d)",
+        context->gl_major_version(), context->gl_minor_version());
+
     glWaitSync(sync_, 0, GL_TIMEOUT_IGNORED);
 #endif
   }
@@ -697,10 +742,13 @@ class GlFenceSyncPoint : public GlSyncPoint {
 
   void Wait() override {
     if (!sync_) return;
-    gl_context_->Run([this] {
-      // TODO: must this run on the original context??
+    if (GlContext::IsAnyContextCurrent()) {
       sync_.Wait();
-    });
+      return;
+    }
+    // In case a current GL context is not available, we fall back using the
+    // captured gl_context_.
+    gl_context_->Run([this] { sync_.Wait(); });
   }
 
   void WaitOnGpu() override {
@@ -812,15 +860,25 @@ class GlNopSyncPoint : public GlSyncPoint {
 #endif
 
 bool GlContext::ShouldUseFenceSync() const {
-#ifdef __EMSCRIPTEN__
+  using internal_gl_context::OpenGlVersion;
+#if defined(__EMSCRIPTEN__)
   // In Emscripten the glWaitSync function is non-null depending on linkopts,
-  // but only works in a WebGL2 context, so fall back to use Finish if it is a
-  // WebGL1/ES2 context.
-  // TODO: apply this more generally once b/152794517 is fixed.
-  return gl_major_version() > 2;
+  // but only works in a WebGL2 context.
+  constexpr OpenGlVersion kMinVersionSyncAvaiable = {.major = 3, .minor = 0};
+#elif defined(MEDIAPIPE_MOBILE)
+  // OpenGL ES, glWaitSync is available since 3.0
+  constexpr OpenGlVersion kMinVersionSyncAvaiable = {.major = 3, .minor = 0};
 #else
-  return SymbolAvailable(&glWaitSync);
-#endif  // __EMSCRIPTEN__
+  // TODO: specify major/minor version per remaining platforms.
+  // By default, ignoring major/minor version requirement for backward
+  // compatibility.
+  constexpr OpenGlVersion kMinVersionSyncAvaiable = {.major = 0, .minor = 0};
+#endif
+
+  return SymbolAvailable(&glWaitSync) &&
+         internal_gl_context::IsOpenGlVersionSameOrAbove(
+             {.major = gl_major_version(), .minor = gl_minor_version()},
+             kMinVersionSyncAvaiable);
 }
 
 std::shared_ptr<GlSyncPoint> GlContext::CreateSyncToken() {

mediapipe/gpu/gl_context.h

@@ -71,6 +71,8 @@ typedef std::function<void()> GlVoidFunction;
 typedef std::function<absl::Status()> GlStatusFunction;
 
 class GlContext;
+// TODO: remove after glWaitSync crashes are resolved.
+class GlSyncWrapper;
 
 // Generic interface for synchronizing access to a shared resource from a
 // different context. This is an abstract class to keep users from
@@ -190,8 +192,7 @@ class GlContext : public std::enable_shared_from_this<GlContext> {
   // Like Run, but does not wait.
   void RunWithoutWaiting(GlVoidFunction gl_func);
 
-  // Returns a synchronization token.
-  // This should not be called outside of the GlContext thread.
+  // Returns a synchronization token for this GlContext.
   std::shared_ptr<GlSyncPoint> CreateSyncToken();
 
   // If another part of the framework calls glFinish, it should call this
@@ -330,6 +331,9 @@ class GlContext : public std::enable_shared_from_this<GlContext> {
       SyncTokenTypeForTest type);
 
  private:
+  // TODO: remove after glWaitSync crashes are resolved.
+  friend GlSyncWrapper;
+
   GlContext();
 
   bool ShouldUseFenceSync() const;
@@ -488,6 +492,18 @@ ABSL_DEPRECATED(
 const GlTextureInfo& GlTextureInfoForGpuBufferFormat(GpuBufferFormat format,
                                                      int plane);
 
+namespace internal_gl_context {
+
+struct OpenGlVersion {
+  int major;
+  int minor;
+};
+
+bool IsOpenGlVersionSameOrAbove(const OpenGlVersion& version,
+                                const OpenGlVersion& expected_version);
+
+}  // namespace internal_gl_context
+
 }  // namespace mediapipe
 
 #endif  // MEDIAPIPE_GPU_GL_CONTEXT_H_

mediapipe/gpu/gl_context_webgl.cc

@@ -32,7 +32,7 @@ namespace mediapipe {
 // TODO: Handle webGL "context lost" and "context restored" events.
 GlContext::StatusOrGlContext GlContext::Create(std::nullptr_t nullp,
                                                bool create_thread) {
-  return Create(0, create_thread);
+  return Create(static_cast<EMSCRIPTEN_WEBGL_CONTEXT_HANDLE>(0), create_thread);
 }
 
 GlContext::StatusOrGlContext GlContext::Create(const GlContext& share_context,

mediapipe/gpu/gpu_buffer_to_image_frame_calculator.cc

@@ -105,7 +105,6 @@ absl::Status GpuBufferToImageFrameCalculator::Process(CalculatorContext* cc) {
                                                         helper_.GetGlVersion());
       glReadPixels(0, 0, src.width(), src.height(), info.gl_format,
                    info.gl_type, frame->MutablePixelData());
-      glFlush();
       cc->Outputs().Index(0).Add(frame.release(), cc->InputTimestamp());
       src.Release();
     });

mediapipe/objc/MPPGraph.cc

@@ -18,6 +18,7 @@
 #import <Accelerate/Accelerate.h>
 
 #include <atomic>
+#include <cstdint>
 
 #import "GTMDefines.h"
 #include "absl/memory/memory.h"
@@ -48,7 +49,7 @@
   std::atomic<int32_t> _framesInFlight;
   /// Used as a sequential timestamp for MediaPipe.
   mediapipe::Timestamp _frameTimestamp;
-  int64 _frameNumber;
+  int64_t _frameNumber;
 
   // Graph config modified to expose requested output streams.
   mediapipe::CalculatorGraphConfig _config;
@@ -90,8 +91,8 @@
                                          &callbackInputName,
                                          /*use_std_function=*/true);
   // No matter what ownership qualifiers are put on the pointer,
-  // NewPermanentCallback will still end up with a strong pointer to MPPGraph*.
-  // That is why we use void* instead.
+  // NewPermanentCallback will still end up with a strong pointer to
+  // MPPGraph*. That is why we use void* instead.
   void* wrapperVoid = (__bridge void*)self;
   _inputSidePackets[callbackInputName] =
       mediapipe::MakePacket<std::function<void(const mediapipe::Packet&)>>(

mediapipe/objc/MPPTimestampConverter.cc

@@ -14,6 +14,8 @@
 
 #import "MPPTimestampConverter.h"
 
+#include <cstdint>
+
 @implementation MPPTimestampConverter {
   mediapipe::Timestamp _mediapipeTimestamp;
   mediapipe::Timestamp _lastTimestamp;
@@ -37,7 +39,7 @@
 - (mediapipe::Timestamp)timestampForMediaTime:(CMTime)mediaTime {
   Float64 sampleSeconds =
       CMTIME_IS_VALID(mediaTime) ? CMTimeGetSeconds(mediaTime) : 0;
-  const int64 sampleUsec =
+  const int64_t sampleUsec =
       sampleSeconds * mediapipe::Timestamp::kTimestampUnitsPerSecond;
   _mediapipeTimestamp = mediapipe::Timestamp(sampleUsec) + _timestampOffset;
   if (_mediapipeTimestamp <= _lastTimestamp) {

mediapipe/objc/util.cc

@@ -14,6 +14,8 @@
 
 #include "mediapipe/objc/util.h"
 
+#include <cstdint>
+
 #include "absl/base/macros.h"
 #include "absl/log/absl_check.h"
 #include "absl/log/absl_log.h"
@@ -639,7 +641,7 @@ std::unique_ptr<mediapipe::ImageFrame> CreateImageFrameForCVPixelBuffer(
   } else {
     frame = absl::make_unique<mediapipe::ImageFrame>(
         image_format, width, height, bytes_per_row,
-        reinterpret_cast<uint8*>(base_address), [image_buffer](uint8* x) {
+        reinterpret_cast<uint8_t*>(base_address), [image_buffer](uint8_t* x) {
           CVPixelBufferUnlockBaseAddress(image_buffer,
                                          kCVPixelBufferLock_ReadOnly);
           CVPixelBufferRelease(image_buffer);

mediapipe/python/BUILD

@@ -97,6 +97,7 @@ cc_library(
         "//mediapipe/tasks/cc/vision/face_landmarker:face_landmarker_graph",
         "//mediapipe/tasks/cc/vision/face_stylizer:face_stylizer_graph",
         "//mediapipe/tasks/cc/vision/gesture_recognizer:gesture_recognizer_graph",
+        "//mediapipe/tasks/cc/vision/holistic_landmarker:holistic_landmarker_graph",
         "//mediapipe/tasks/cc/vision/image_classifier:image_classifier_graph",
         "//mediapipe/tasks/cc/vision/image_embedder:image_embedder_graph",
         "//mediapipe/tasks/cc/vision/image_segmenter:image_segmenter_graph",

mediapipe/tasks/c/components/containers/BUILD

@@ -43,6 +43,33 @@ cc_test(
     ],
 )
 
+cc_library(
+    name = "matrix",
+    hdrs = ["matrix.h"],
+)
+
+cc_library(
+    name = "matrix_converter",
+    srcs = ["matrix_converter.cc"],
+    hdrs = ["matrix_converter.h"],
+    deps = [
+        ":matrix",
+        "@eigen_archive//:eigen3",
+    ],
+)
+
+cc_test(
+    name = "matrix_converter_test",
+    srcs = ["matrix_converter_test.cc"],
+    deps = [
+        ":matrix",
+        ":matrix_converter",
+        "//mediapipe/framework/port:gtest",
+        "@com_google_googletest//:gtest_main",
+        "@eigen_archive//:eigen3",
+    ],
+)
+
 cc_library(
     name = "landmark",
     hdrs = ["landmark.h"],

mediapipe/tasks/c/components/containers/matrix.h

@@ -0,0 +1,41 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#ifndef MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_MATRIX_H_
+#define MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_MATRIX_H_
+
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Data are stored in column-major order by default.
+struct Matrix {
+  // The number of rows in the matrix.
+  uint32_t rows;
+
+  // The number of columns in the matrix.
+  uint32_t cols;
+
+  // The matrix data stored in a column-first layout.
+  float* data;
+};
+
+#ifdef __cplusplus
+}  // extern C
+#endif
+
+#endif  // MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_MATRIX_H_

mediapipe/tasks/c/components/containers/matrix_converter.cc

@@ -0,0 +1,43 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#include "mediapipe/tasks/c/components/containers/matrix_converter.h"
+
+#include <cstring>
+
+#include "Eigen/Core"
+#include "mediapipe/tasks/c/components/containers/matrix.h"
+
+namespace mediapipe::tasks::c::components::containers {
+
+void CppConvertToMatrix(const Eigen::MatrixXf& in, ::Matrix* out) {
+  out->rows = in.rows();
+  out->cols = in.cols();
+  out->data = new float[out->rows * out->cols];
+
+  // Copies data from an Eigen matrix (default column-major as used by
+  // MediaPipe) to a C-style matrix, preserving the sequence of elements as per
+  // the Eigen matrix's internal storage (column-major order by default).
+  memcpy(out->data, in.data(), sizeof(float) * out->rows * out->cols);
+}
+
+void CppCloseMatrix(::Matrix* m) {
+  if (m->data) {
+    delete[] m->data;
+    m->data = nullptr;
+  }
+}
+
+}  // namespace mediapipe::tasks::c::components::containers

mediapipe/tasks/c/components/containers/matrix_converter.h

@@ -0,0 +1,30 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#ifndef MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_MATRIX_CONVERTER_H_
+#define MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_MATRIX_CONVERTER_H_
+
+#include "Eigen/Core"
+#include "mediapipe/tasks/c/components/containers/matrix.h"
+
+namespace mediapipe::tasks::c::components::containers {
+
+void CppConvertToMatrix(const Eigen::MatrixXf& in, ::Matrix* out);
+
+void CppCloseMatrix(::Matrix* m);
+
+}  // namespace mediapipe::tasks::c::components::containers
+
+#endif  // MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_MATRIX_CONVERTER_H_

mediapipe/tasks/c/components/containers/matrix_converter_test.cc

@@ -0,0 +1,49 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#include "mediapipe/tasks/c/components/containers/matrix_converter.h"
+
+#include "Eigen/Core"
+#include "mediapipe/framework/port/gtest.h"
+#include "mediapipe/tasks/c/components/containers/matrix.h"
+
+namespace mediapipe::tasks::c::components::containers {
+
+TEST(MatrixConversionTest, ConvertsEigenMatrixToCMatrixAndFreesMemory) {
+  // Initialize an Eigen::MatrixXf
+  Eigen::MatrixXf cpp_matrix(2, 2);
+  cpp_matrix << 1.0f, 2.0f, 3.0f, 4.0f;
+
+  // Convert this Eigen matrix to C-style Matrix
+  ::Matrix c_matrix;
+  CppConvertToMatrix(cpp_matrix, &c_matrix);
+
+  // Verify the conversion
+  EXPECT_EQ(c_matrix.rows, 2);
+  EXPECT_EQ(c_matrix.cols, 2);
+  ASSERT_NE(c_matrix.data, nullptr);
+  EXPECT_FLOAT_EQ(c_matrix.data[0], 1.0f);
+  EXPECT_FLOAT_EQ(c_matrix.data[1], 3.0f);
+  EXPECT_FLOAT_EQ(c_matrix.data[2], 2.0f);
+  EXPECT_FLOAT_EQ(c_matrix.data[3], 4.0f);
+
+  // Close the C-style Matrix
+  CppCloseMatrix(&c_matrix);
+
+  // Verify that memory is freed
+  EXPECT_EQ(c_matrix.data, nullptr);
+}
+
+}  // namespace mediapipe::tasks::c::components::containers

mediapipe/tasks/c/vision/face_landmarker/BUILD

@@ -0,0 +1,149 @@
+# Copyright 2023 The MediaPipe Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+package(default_visibility = ["//mediapipe/tasks:internal"])
+
+licenses(["notice"])
+
+cc_library(
+    name = "face_landmarker_result",
+    hdrs = ["face_landmarker_result.h"],
+    visibility = ["//visibility:public"],
+    deps = [
+        "//mediapipe/tasks/c/components/containers:category",
+        "//mediapipe/tasks/c/components/containers:landmark",
+        "//mediapipe/tasks/c/components/containers:matrix",
+    ],
+)
+
+cc_library(
+    name = "face_landmarker_result_converter",
+    srcs = ["face_landmarker_result_converter.cc"],
+    hdrs = ["face_landmarker_result_converter.h"],
+    deps = [
+        ":face_landmarker_result",
+        "//mediapipe/tasks/c/components/containers:category",
+        "//mediapipe/tasks/c/components/containers:category_converter",
+        "//mediapipe/tasks/c/components/containers:landmark",
+        "//mediapipe/tasks/c/components/containers:landmark_converter",
+        "//mediapipe/tasks/c/components/containers:matrix",
+        "//mediapipe/tasks/c/components/containers:matrix_converter",
+        "//mediapipe/tasks/cc/components/containers:category",
+        "//mediapipe/tasks/cc/components/containers:landmark",
+        "//mediapipe/tasks/cc/vision/face_landmarker:face_landmarker_result",
+    ],
+)
+
+cc_test(
+    name = "face_landmarker_result_converter_test",
+    srcs = ["face_landmarker_result_converter_test.cc"],
+    linkstatic = 1,
+    deps = [
+        ":face_landmarker_result",
+        ":face_landmarker_result_converter",
+        "//mediapipe/framework/formats:matrix",
+        "//mediapipe/framework/port:gtest",
+        "//mediapipe/tasks/c/components/containers:landmark",
+        "//mediapipe/tasks/cc/components/containers:category",
+        "//mediapipe/tasks/cc/components/containers:classification_result",
+        "//mediapipe/tasks/cc/components/containers:landmark",
+        "//mediapipe/tasks/cc/vision/face_landmarker:face_landmarker_result",
+        "@com_google_googletest//:gtest_main",
+        "@eigen_archive//:eigen3",
+    ],
+)
+
+cc_library(
+    name = "face_landmarker_lib",
+    srcs = ["face_landmarker.cc"],
+    hdrs = ["face_landmarker.h"],
+    visibility = ["//visibility:public"],
+    deps = [
+        ":face_landmarker_result",
+        ":face_landmarker_result_converter",
+        "//mediapipe/framework/formats:image",
+        "//mediapipe/framework/formats:image_frame",
+        "//mediapipe/tasks/c/core:base_options",
+        "//mediapipe/tasks/c/core:base_options_converter",
+        "//mediapipe/tasks/c/vision/core:common",
+        "//mediapipe/tasks/cc/vision/core:running_mode",
+        "//mediapipe/tasks/cc/vision/face_landmarker",
+        "//mediapipe/tasks/cc/vision/face_landmarker:face_landmarker_result",
+        "//mediapipe/tasks/cc/vision/utils:image_utils",
+        "@com_google_absl//absl/log:absl_log",
+        "@com_google_absl//absl/status",
+        "@com_google_absl//absl/status:statusor",
+    ],
+    alwayslink = 1,
+)
+
+cc_test(
+    name = "face_landmarker_test",
+    srcs = ["face_landmarker_test.cc"],
+    data = [
+        "//mediapipe/framework/formats:image_frame_opencv",
+        "//mediapipe/framework/port:opencv_core",
+        "//mediapipe/framework/port:opencv_imgproc",
+        "//mediapipe/tasks/testdata/vision:test_images",
+        "//mediapipe/tasks/testdata/vision:test_models",
+    ],
+    linkstatic = 1,
+    deps = [
+        ":face_landmarker_lib",
+        ":face_landmarker_result",
+        "//mediapipe/framework/deps:file_path",
+        "//mediapipe/framework/formats:image",
+        "//mediapipe/framework/port:gtest",
+        "//mediapipe/tasks/c/components/containers:landmark",
+        "//mediapipe/tasks/c/vision/core:common",
+        "//mediapipe/tasks/cc/vision/utils:image_utils",
+        "@com_google_absl//absl/flags:flag",
+        "@com_google_absl//absl/strings",
+        "@com_google_googletest//:gtest_main",
+    ],
+)
+
+# bazel build -c opt --linkopt -s --strip always --define MEDIAPIPE_DISABLE_GPU=1 \
+# //mediapipe/tasks/c/vision/face_landmarker:libface_landmarker.so
+cc_binary(
+    name = "libface_landmarker.so",
+    linkopts = [
+        "-Wl,-soname=libface_landmarker.so",
+        "-fvisibility=hidden",
+    ],
+    linkshared = True,
+    tags = [
+        "manual",
+        "nobuilder",
+        "notap",
+    ],
+    deps = [":face_landmarker_lib"],
+)
+
+# bazel build --config darwin_arm64 -c opt --strip always --define MEDIAPIPE_DISABLE_GPU=1 \
+# //mediapipe/tasks/c/vision/face_landmarker:libface_landmarker.dylib
+cc_binary(
+    name = "libface_landmarker.dylib",
+    linkopts = [
+        "-Wl,-install_name,libface_landmarker.dylib",
+        "-fvisibility=hidden",
+    ],
+    linkshared = True,
+    tags = [
+        "manual",
+        "nobuilder",
+        "notap",
+    ],
+    deps = [":face_landmarker_lib"],
+)

mediapipe/tasks/c/vision/face_landmarker/face_landmarker.cc

@@ -0,0 +1,287 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker.h"
+
+#include <cstdint>
+#include <cstdlib>
+#include <memory>
+#include <utility>
+
+#include "absl/log/absl_log.h"
+#include "absl/status/status.h"
+#include "absl/status/statusor.h"
+#include "mediapipe/framework/formats/image.h"
+#include "mediapipe/framework/formats/image_frame.h"
+#include "mediapipe/tasks/c/core/base_options_converter.h"
+#include "mediapipe/tasks/c/vision/core/common.h"
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result.h"
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result_converter.h"
+#include "mediapipe/tasks/cc/vision/core/running_mode.h"
+#include "mediapipe/tasks/cc/vision/face_landmarker/face_landmarker.h"
+#include "mediapipe/tasks/cc/vision/face_landmarker/face_landmarker_result.h"
+#include "mediapipe/tasks/cc/vision/utils/image_utils.h"
+
+namespace mediapipe::tasks::c::vision::face_landmarker {
+
+namespace {
+
+using ::mediapipe::tasks::c::components::containers::
+    CppCloseFaceLandmarkerResult;
+using ::mediapipe::tasks::c::components::containers::
+    CppConvertToFaceLandmarkerResult;
+using ::mediapipe::tasks::c::core::CppConvertToBaseOptions;
+using ::mediapipe::tasks::vision::CreateImageFromBuffer;
+using ::mediapipe::tasks::vision::core::RunningMode;
+using ::mediapipe::tasks::vision::face_landmarker::FaceLandmarker;
+typedef ::mediapipe::tasks::vision::face_landmarker::FaceLandmarkerResult
+    CppFaceLandmarkerResult;
+
+int CppProcessError(absl::Status status, char** error_msg) {
+  if (error_msg) {
+    *error_msg = strdup(status.ToString().c_str());
+  }
+  return status.raw_code();
+}
+
+}  // namespace
+
+void CppConvertToFaceLandmarkerOptions(
+    const FaceLandmarkerOptions& in,
+    mediapipe::tasks::vision::face_landmarker::FaceLandmarkerOptions* out) {
+  out->num_faces = in.num_faces;
+  out->min_face_detection_confidence = in.min_face_detection_confidence;
+  out->min_face_presence_confidence = in.min_face_presence_confidence;
+  out->min_tracking_confidence = in.min_tracking_confidence;
+  out->output_face_blendshapes = in.output_face_blendshapes;
+  out->output_facial_transformation_matrixes =
+      in.output_facial_transformation_matrixes;
+}
+
+FaceLandmarker* CppFaceLandmarkerCreate(const FaceLandmarkerOptions& options,
+                                        char** error_msg) {
+  auto cpp_options = std::make_unique<
+      ::mediapipe::tasks::vision::face_landmarker::FaceLandmarkerOptions>();
+
+  CppConvertToBaseOptions(options.base_options, &cpp_options->base_options);
+  CppConvertToFaceLandmarkerOptions(options, cpp_options.get());
+  cpp_options->running_mode = static_cast<RunningMode>(options.running_mode);
+
+  // Enable callback for processing live stream data when the running mode is
+  // set to RunningMode::LIVE_STREAM.
+  if (cpp_options->running_mode == RunningMode::LIVE_STREAM) {
+    if (options.result_callback == nullptr) {
+      const absl::Status status = absl::InvalidArgumentError(
+          "Provided null pointer to callback function.");
+      ABSL_LOG(ERROR) << "Failed to create FaceLandmarker: " << status;
+      CppProcessError(status, error_msg);
+      return nullptr;
+    }
+
+    FaceLandmarkerOptions::result_callback_fn result_callback =
+        options.result_callback;
+    cpp_options->result_callback =
+        [result_callback](absl::StatusOr<CppFaceLandmarkerResult> cpp_result,
+                          const Image& image, int64_t timestamp) {
+          char* error_msg = nullptr;
+
+          if (!cpp_result.ok()) {
+            ABSL_LOG(ERROR) << "Detection failed: " << cpp_result.status();
+            CppProcessError(cpp_result.status(), &error_msg);
+            result_callback({}, MpImage(), timestamp, error_msg);
+            free(error_msg);
+            return;
+          }
+
+          // Result is valid for the lifetime of the callback function.
+          FaceLandmarkerResult result;
+          CppConvertToFaceLandmarkerResult(*cpp_result, &result);
+
+          const auto& image_frame = image.GetImageFrameSharedPtr();
+          const MpImage mp_image = {
+              .type = MpImage::IMAGE_FRAME,
+              .image_frame = {
+                  .format = static_cast<::ImageFormat>(image_frame->Format()),
+                  .image_buffer = image_frame->PixelData(),
+                  .width = image_frame->Width(),
+                  .height = image_frame->Height()}};
+
+          result_callback(&result, mp_image, timestamp,
+                          /* error_msg= */ nullptr);
+
+          CppCloseFaceLandmarkerResult(&result);
+        };
+  }
+
+  auto landmarker = FaceLandmarker::Create(std::move(cpp_options));
+  if (!landmarker.ok()) {
+    ABSL_LOG(ERROR) << "Failed to create FaceLandmarker: "
+                    << landmarker.status();
+    CppProcessError(landmarker.status(), error_msg);
+    return nullptr;
+  }
+  return landmarker->release();
+}
+
+int CppFaceLandmarkerDetect(void* landmarker, const MpImage& image,
+                            FaceLandmarkerResult* result, char** error_msg) {
+  if (image.type == MpImage::GPU_BUFFER) {
+    const absl::Status status =
+        absl::InvalidArgumentError("GPU Buffer not supported yet.");
+
+    ABSL_LOG(ERROR) << "Detection failed: " << status.message();
+    return CppProcessError(status, error_msg);
+  }
+
+  const auto img = CreateImageFromBuffer(
+      static_cast<ImageFormat::Format>(image.image_frame.format),
+      image.image_frame.image_buffer, image.image_frame.width,
+      image.image_frame.height);
+
+  if (!img.ok()) {
+    ABSL_LOG(ERROR) << "Failed to create Image: " << img.status();
+    return CppProcessError(img.status(), error_msg);
+  }
+
+  auto cpp_landmarker = static_cast<FaceLandmarker*>(landmarker);
+  auto cpp_result = cpp_landmarker->Detect(*img);
+  if (!cpp_result.ok()) {
+    ABSL_LOG(ERROR) << "Detection failed: " << cpp_result.status();
+    return CppProcessError(cpp_result.status(), error_msg);
+  }
+  CppConvertToFaceLandmarkerResult(*cpp_result, result);
+  return 0;
+}
+
+int CppFaceLandmarkerDetectForVideo(void* landmarker, const MpImage& image,
+                                    int64_t timestamp_ms,
+                                    FaceLandmarkerResult* result,
+                                    char** error_msg) {
+  if (image.type == MpImage::GPU_BUFFER) {
+    absl::Status status =
+        absl::InvalidArgumentError("GPU Buffer not supported yet");
+
+    ABSL_LOG(ERROR) << "Detection failed: " << status.message();
+    return CppProcessError(status, error_msg);
+  }
+
+  const auto img = CreateImageFromBuffer(
+      static_cast<ImageFormat::Format>(image.image_frame.format),
+      image.image_frame.image_buffer, image.image_frame.width,
+      image.image_frame.height);
+
+  if (!img.ok()) {
+    ABSL_LOG(ERROR) << "Failed to create Image: " << img.status();
+    return CppProcessError(img.status(), error_msg);
+  }
+
+  auto cpp_landmarker = static_cast<FaceLandmarker*>(landmarker);
+  auto cpp_result = cpp_landmarker->DetectForVideo(*img, timestamp_ms);
+  if (!cpp_result.ok()) {
+    ABSL_LOG(ERROR) << "Detection failed: " << cpp_result.status();
+    return CppProcessError(cpp_result.status(), error_msg);
+  }
+  CppConvertToFaceLandmarkerResult(*cpp_result, result);
+  return 0;
+}
+
+int CppFaceLandmarkerDetectAsync(void* landmarker, const MpImage& image,
+                                 int64_t timestamp_ms, char** error_msg) {
+  if (image.type == MpImage::GPU_BUFFER) {
+    absl::Status status =
+        absl::InvalidArgumentError("GPU Buffer not supported yet");
+
+    ABSL_LOG(ERROR) << "Detection failed: " << status.message();
+    return CppProcessError(status, error_msg);
+  }
+
+  const auto img = CreateImageFromBuffer(
+      static_cast<ImageFormat::Format>(image.image_frame.format),
+      image.image_frame.image_buffer, image.image_frame.width,
+      image.image_frame.height);
+
+  if (!img.ok()) {
+    ABSL_LOG(ERROR) << "Failed to create Image: " << img.status();
+    return CppProcessError(img.status(), error_msg);
+  }
+
+  auto cpp_landmarker = static_cast<FaceLandmarker*>(landmarker);
+  auto cpp_result = cpp_landmarker->DetectAsync(*img, timestamp_ms);
+  if (!cpp_result.ok()) {
+    ABSL_LOG(ERROR) << "Data preparation for the landmark detection failed: "
+                    << cpp_result;
+    return CppProcessError(cpp_result, error_msg);
+  }
+  return 0;
+}
+
+void CppFaceLandmarkerCloseResult(FaceLandmarkerResult* result) {
+  CppCloseFaceLandmarkerResult(result);
+}
+
+int CppFaceLandmarkerClose(void* landmarker, char** error_msg) {
+  auto cpp_landmarker = static_cast<FaceLandmarker*>(landmarker);
+  auto result = cpp_landmarker->Close();
+  if (!result.ok()) {
+    ABSL_LOG(ERROR) << "Failed to close FaceLandmarker: " << result;
+    return CppProcessError(result, error_msg);
+  }
+  delete cpp_landmarker;
+  return 0;
+}
+
+}  // namespace mediapipe::tasks::c::vision::face_landmarker
+
+extern "C" {
+
+void* face_landmarker_create(struct FaceLandmarkerOptions* options,
+                             char** error_msg) {
+  return mediapipe::tasks::c::vision::face_landmarker::CppFaceLandmarkerCreate(
+      *options, error_msg);
+}
+
+int face_landmarker_detect_image(void* landmarker, const MpImage& image,
+                                 FaceLandmarkerResult* result,
+                                 char** error_msg) {
+  return mediapipe::tasks::c::vision::face_landmarker::CppFaceLandmarkerDetect(
+      landmarker, image, result, error_msg);
+}
+
+int face_landmarker_detect_for_video(void* landmarker, const MpImage& image,
+                                     int64_t timestamp_ms,
+                                     FaceLandmarkerResult* result,
+                                     char** error_msg) {
+  return mediapipe::tasks::c::vision::face_landmarker::
+      CppFaceLandmarkerDetectForVideo(landmarker, image, timestamp_ms, result,
+                                      error_msg);
+}
+
+int face_landmarker_detect_async(void* landmarker, const MpImage& image,
+                                 int64_t timestamp_ms, char** error_msg) {
+  return mediapipe::tasks::c::vision::face_landmarker::
+      CppFaceLandmarkerDetectAsync(landmarker, image, timestamp_ms, error_msg);
+}
+
+void face_landmarker_close_result(FaceLandmarkerResult* result) {
+  mediapipe::tasks::c::vision::face_landmarker::CppFaceLandmarkerCloseResult(
+      result);
+}
+
+int face_landmarker_close(void* landmarker, char** error_ms) {
+  return mediapipe::tasks::c::vision::face_landmarker::CppFaceLandmarkerClose(
+      landmarker, error_ms);
+}
+
+}  // extern "C"

mediapipe/tasks/c/vision/face_landmarker/face_landmarker.h

@@ -0,0 +1,156 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#ifndef MEDIAPIPE_TASKS_C_VISION_FACE_LANDMARKER_FACE_LANDMARKER_H_
+#define MEDIAPIPE_TASKS_C_VISION_FACE_LANDMARKER_FACE_LANDMARKER_H_
+
+#include "mediapipe/tasks/c/core/base_options.h"
+#include "mediapipe/tasks/c/vision/core/common.h"
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result.h"
+
+#ifndef MP_EXPORT
+#define MP_EXPORT __attribute__((visibility("default")))
+#endif  // MP_EXPORT
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// The options for configuring a MediaPipe face landmarker task.
+struct FaceLandmarkerOptions {
+  // Base options for configuring MediaPipe Tasks, such as specifying the model
+  // file with metadata, accelerator options, op resolver, etc.
+  struct BaseOptions base_options;
+
+  // The running mode of the task. Default to the image mode.
+  // FaceLandmarker has three running modes:
+  // 1) The image mode for recognizing face landmarks on single image inputs.
+  // 2) The video mode for recognizing face landmarks on the decoded frames of a
+  //    video.
+  // 3) The live stream mode for recognizing face landmarks on the live stream
+  //    of input data, such as from camera. In this mode, the "result_callback"
+  //    below must be specified to receive the detection results asynchronously.
+  RunningMode running_mode;
+
+  // The maximum number of faces can be detected by the FaceLandmarker.
+  int num_faces = 1;
+
+  // The minimum confidence score for the face detection to be considered
+  // successful.
+  float min_face_detection_confidence = 0.5;
+
+  // The minimum confidence score of face presence score in the face landmark
+  // detection.
+  float min_face_presence_confidence = 0.5;
+
+  // The minimum confidence score for the face tracking to be considered
+  // successful.
+  float min_tracking_confidence = 0.5;
+
+  // Whether FaceLandmarker outputs face blendshapes classification. Face
+  // blendshapes are used for rendering the 3D face model.
+  bool output_face_blendshapes = false;
+
+  // Whether FaceLandmarker outputs facial transformation_matrix. Facial
+  // transformation matrix is used to transform the face landmarks in canonical
+  // face to the detected face, so that users can apply face effects on the
+  // detected landmarks.
+  bool output_facial_transformation_matrixes = false;
+
+  // The user-defined result callback for processing live stream data.
+  // The result callback should only be specified when the running mode is set
+  // to RunningMode::LIVE_STREAM. Arguments of the callback function include:
+  // the pointer to recognition result, the image that result was obtained
+  // on, the timestamp relevant to recognition results and pointer to error
+  // message in case of any failure. The validity of the passed arguments is
+  // true for the lifetime of the callback function.
+  //
+  // A caller is responsible for closing face landmarker result.
+  typedef void (*result_callback_fn)(const FaceLandmarkerResult* result,
+                                     const MpImage& image, int64_t timestamp_ms,
+                                     char* error_msg);
+  result_callback_fn result_callback;
+};
+
+// Creates an FaceLandmarker from the provided `options`.
+// Returns a pointer to the face landmarker on success.
+// If an error occurs, returns `nullptr` and sets the error parameter to an
+// an error message (if `error_msg` is not `nullptr`). You must free the memory
+// allocated for the error message.
+MP_EXPORT void* face_landmarker_create(struct FaceLandmarkerOptions* options,
+                                       char** error_msg);
+
+// Performs face landmark detection on the input `image`. Returns `0` on
+// success. If an error occurs, returns an error code and sets the error
+// parameter to an an error message (if `error_msg` is not `nullptr`). You must
+// free the memory allocated for the error message.
+MP_EXPORT int face_landmarker_detect_image(void* landmarker,
+                                           const MpImage& image,
+                                           FaceLandmarkerResult* result,
+                                           char** error_msg);
+
+// Performs face landmark detection on the provided video frame.
+// Only use this method when the FaceLandmarker is created with the video
+// running mode.
+// The image can be of any size with format RGB or RGBA. It's required to
+// provide the video frame's timestamp (in milliseconds). The input timestamps
+// must be monotonically increasing.
+// If an error occurs, returns an error code and sets the error parameter to an
+// an error message (if `error_msg` is not `nullptr`). You must free the memory
+// allocated for the error message.
+MP_EXPORT int face_landmarker_detect_for_video(void* landmarker,
+                                               const MpImage& image,
+                                               int64_t timestamp_ms,
+                                               FaceLandmarkerResult* result,
+                                               char** error_msg);
+
+// Sends live image data to face landmark detection, and the results will be
+// available via the `result_callback` provided in the FaceLandmarkerOptions.
+// Only use this method when the FaceLandmarker is created with the live
+// stream running mode.
+// The image can be of any size with format RGB or RGBA. It's required to
+// provide a timestamp (in milliseconds) to indicate when the input image is
+// sent to the face landmarker. The input timestamps must be monotonically
+// increasing.
+// The `result_callback` provides:
+//   - The recognition results as an FaceLandmarkerResult object.
+//   - The const reference to the corresponding input image that the face
+//     landmarker runs on. Note that the const reference to the image will no
+//     longer be valid when the callback returns. To access the image data
+//     outside of the callback, callers need to make a copy of the image.
+//   - The input timestamp in milliseconds.
+// If an error occurs, returns an error code and sets the error parameter to an
+// an error message (if `error_msg` is not `nullptr`). You must free the memory
+// allocated for the error message.
+MP_EXPORT int face_landmarker_detect_async(void* landmarker,
+                                           const MpImage& image,
+                                           int64_t timestamp_ms,
+                                           char** error_msg);
+
+// Frees the memory allocated inside a FaceLandmarkerResult result.
+// Does not free the result pointer itself.
+MP_EXPORT void face_landmarker_close_result(FaceLandmarkerResult* result);
+
+// Frees face landmarker.
+// If an error occurs, returns an error code and sets the error parameter to an
+// an error message (if `error_msg` is not `nullptr`). You must free the memory
+// allocated for the error message.
+MP_EXPORT int face_landmarker_close(void* landmarker, char** error_msg);
+
+#ifdef __cplusplus
+}  // extern C
+#endif
+
+#endif  // MEDIAPIPE_TASKS_C_VISION_FACE_LANDMARKER_FACE_LANDMARKER_H_

mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result.h

@@ -0,0 +1,59 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#ifndef MEDIAPIPE_TASKS_C_VISION_FACE_LANDMARKER_RESULT_FACE_LANDMARKER_RESULT_H_
+#define MEDIAPIPE_TASKS_C_VISION_FACE_LANDMARKER_RESULT_FACE_LANDMARKER_RESULT_H_
+
+#include <cstdint>
+
+#include "mediapipe/tasks/c/components/containers/category.h"
+#include "mediapipe/tasks/c/components/containers/landmark.h"
+#include "mediapipe/tasks/c/components/containers/matrix.h"
+
+#ifndef MP_EXPORT
+#define MP_EXPORT __attribute__((visibility("default")))
+#endif  // MP_EXPORT
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// The hand landmarker result from HandLandmarker, where each vector
+// element represents a single hand detected in the image.
+struct FaceLandmarkerResult {
+  // Detected face landmarks in normalized image coordinates.
+  struct NormalizedLandmarks* face_landmarks;
+
+  // The number of elements in the face_landmarks array.
+  uint32_t face_landmarks_count;
+
+  // Optional face blendshapes results.
+  struct Categories* face_blendshapes;
+
+  // The number of elements in the face_blendshapes array.
+  uint32_t face_blendshapes_count;
+
+  // Optional facial transformation matrixes.
+  struct Matrix* facial_transformation_matrixes;
+
+  // The number of elements in the facial_transformation_matrixes array.
+  uint32_t facial_transformation_matrixes_count;
+};
+
+#ifdef __cplusplus
+}  // extern C
+#endif
+
+#endif  // MEDIAPIPE_TASKS_C_VISION_FACE_LANDMARKER_RESULT_FACE_LANDMARKER_RESULT_H_

mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result_converter.cc

@@ -0,0 +1,117 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result_converter.h"
+
+#include <cstdint>
+#include <vector>
+
+#include "mediapipe/tasks/c/components/containers/category.h"
+#include "mediapipe/tasks/c/components/containers/category_converter.h"
+#include "mediapipe/tasks/c/components/containers/landmark.h"
+#include "mediapipe/tasks/c/components/containers/landmark_converter.h"
+#include "mediapipe/tasks/c/components/containers/matrix.h"
+#include "mediapipe/tasks/c/components/containers/matrix_converter.h"
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result.h"
+#include "mediapipe/tasks/cc/components/containers/category.h"
+#include "mediapipe/tasks/cc/components/containers/landmark.h"
+#include "mediapipe/tasks/cc/vision/face_landmarker/face_landmarker_result.h"
+
+namespace mediapipe::tasks::c::components::containers {
+
+using CppCategory = ::mediapipe::tasks::components::containers::Category;
+using CppNormalizedLandmark =
+    ::mediapipe::tasks::components::containers::NormalizedLandmark;
+
+void CppConvertToFaceLandmarkerResult(
+    const ::mediapipe::tasks::vision::face_landmarker::FaceLandmarkerResult& in,
+    FaceLandmarkerResult* out) {
+  out->face_landmarks_count = in.face_landmarks.size();
+  out->face_landmarks = new NormalizedLandmarks[out->face_landmarks_count];
+  for (uint32_t i = 0; i < out->face_landmarks_count; ++i) {
+    std::vector<CppNormalizedLandmark> cpp_normalized_landmarks;
+    for (uint32_t j = 0; j < in.face_landmarks[i].landmarks.size(); ++j) {
+      const auto& cpp_landmark = in.face_landmarks[i].landmarks[j];
+      cpp_normalized_landmarks.push_back(cpp_landmark);
+    }
+    CppConvertToNormalizedLandmarks(cpp_normalized_landmarks,
+                                    &out->face_landmarks[i]);
+  }
+
+  if (in.face_blendshapes.has_value()) {
+    out->face_blendshapes_count = in.face_blendshapes->size();
+    out->face_blendshapes = new Categories[out->face_blendshapes_count];
+
+    for (uint32_t i = 0; i < out->face_blendshapes_count; ++i) {
+      uint32_t categories_count =
+          in.face_blendshapes.value()[i].categories.size();
+      out->face_blendshapes[i].categories_count = categories_count;
+      out->face_blendshapes[i].categories = new Category[categories_count];
+
+      for (uint32_t j = 0; j < categories_count; ++j) {
+        const auto& cpp_category = in.face_blendshapes.value()[i].categories[j];
+        CppConvertToCategory(cpp_category,
+                             &out->face_blendshapes[i].categories[j]);
+      }
+    }
+  } else {
+    out->face_blendshapes_count = 0;
+    out->face_blendshapes = nullptr;
+  }
+
+  if (in.facial_transformation_matrixes.has_value()) {
+    out->facial_transformation_matrixes_count =
+        in.facial_transformation_matrixes.value().size();
+    out->facial_transformation_matrixes =
+        new ::Matrix[out->facial_transformation_matrixes_count];
+    for (uint32_t i = 0; i < out->facial_transformation_matrixes_count; ++i) {
+      CppConvertToMatrix(in.facial_transformation_matrixes.value()[i],
+                         &out->facial_transformation_matrixes[i]);
+    }
+  } else {
+    out->facial_transformation_matrixes_count = 0;
+    out->facial_transformation_matrixes = nullptr;
+  }
+}
+
+void CppCloseFaceLandmarkerResult(FaceLandmarkerResult* result) {
+  for (uint32_t i = 0; i < result->face_blendshapes_count; ++i) {
+    for (uint32_t j = 0; j < result->face_blendshapes[i].categories_count;
+         ++j) {
+      CppCloseCategory(&result->face_blendshapes[i].categories[j]);
+    }
+    delete[] result->face_blendshapes[i].categories;
+  }
+  delete[] result->face_blendshapes;
+
+  for (uint32_t i = 0; i < result->face_landmarks_count; ++i) {
+    CppCloseNormalizedLandmarks(&result->face_landmarks[i]);
+  }
+  delete[] result->face_landmarks;
+
+  for (uint32_t i = 0; i < result->facial_transformation_matrixes_count; ++i) {
+    CppCloseMatrix(&result->facial_transformation_matrixes[i]);
+  }
+  delete[] result->facial_transformation_matrixes;
+
+  result->face_blendshapes_count = 0;
+  result->face_landmarks_count = 0;
+  result->facial_transformation_matrixes_count = 0;
+  result->face_blendshapes = nullptr;
+  result->face_landmarks = nullptr;
+  result->facial_transformation_matrixes = nullptr;
+}
+
+}  // namespace mediapipe::tasks::c::components::containers

mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result_converter.h

@@ -0,0 +1,32 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#ifndef MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_FACE_LANDMARKER_RESULT_CONVERTER_H_
+#define MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_FACE_LANDMARKER_RESULT_CONVERTER_H_
+
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result.h"
+#include "mediapipe/tasks/cc/vision/face_landmarker/face_landmarker_result.h"
+
+namespace mediapipe::tasks::c::components::containers {
+
+void CppConvertToFaceLandmarkerResult(
+    const mediapipe::tasks::vision::face_landmarker::FaceLandmarkerResult& in,
+    FaceLandmarkerResult* out);
+
+void CppCloseFaceLandmarkerResult(FaceLandmarkerResult* result);
+
+}  // namespace mediapipe::tasks::c::components::containers
+
+#endif  // MEDIAPIPE_TASKS_C_COMPONENTS_CONTAINERS_FACE_LANDMARKER_RESULT_CONVERTER_H_

mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result_converter_test.cc

@@ -0,0 +1,154 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result_converter.h"
+
+#include <cstddef>
+#include <cstdint>
+#include <string>
+#include <vector>
+
+#include "Eigen/Core"
+#include "mediapipe/framework/formats/matrix.h"
+#include "mediapipe/framework/port/gtest.h"
+#include "mediapipe/tasks/c/components/containers/landmark.h"
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result.h"
+#include "mediapipe/tasks/cc/components/containers/category.h"
+#include "mediapipe/tasks/cc/components/containers/classification_result.h"
+#include "mediapipe/tasks/cc/components/containers/landmark.h"
+#include "mediapipe/tasks/cc/vision/face_landmarker/face_landmarker_result.h"
+
+namespace mediapipe::tasks::c::components::containers {
+
+void InitFaceLandmarkerResult(
+    ::mediapipe::tasks::vision::face_landmarker::FaceLandmarkerResult*
+        cpp_result) {
+  // Initialize face_landmarks
+  mediapipe::tasks::components::containers::NormalizedLandmark
+      cpp_normalized_landmark = {/* x= */ 0.1f, /* y= */ 0.2f, /* z= */ 0.3f};
+  mediapipe::tasks::components::containers::NormalizedLandmarks
+      cpp_normalized_landmarks;
+  cpp_normalized_landmarks.landmarks.push_back(cpp_normalized_landmark);
+  cpp_result->face_landmarks.push_back(cpp_normalized_landmarks);
+
+  // Initialize face_blendshapes
+  mediapipe::tasks::components::containers::Category cpp_category = {
+      /* index= */ 1,
+      /* score= */ 0.8f,
+      /* category_name= */ "blendshape_label_1",
+      /* display_name= */ "blendshape_display_name_1"};
+  mediapipe::tasks::components::containers::Classifications
+      classifications_for_blendshapes;
+  classifications_for_blendshapes.categories.push_back(cpp_category);
+
+  cpp_result->face_blendshapes =
+      std::vector<mediapipe::tasks::components::containers::Classifications>{
+          classifications_for_blendshapes};
+  cpp_result->face_blendshapes->push_back(classifications_for_blendshapes);
+
+  // Initialize facial_transformation_matrixes
+  Eigen::MatrixXf cpp_matrix(2, 2);
+  cpp_matrix << 1.0f, 2.0f, 3.0f, 4.0f;
+  cpp_result->facial_transformation_matrixes = std::vector<Matrix>{cpp_matrix};
+}
+
+TEST(FaceLandmarkerResultConverterTest, ConvertsCustomResult) {
+  // Initialize a C++ FaceLandmarkerResult
+  ::mediapipe::tasks::vision::face_landmarker::FaceLandmarkerResult cpp_result;
+  InitFaceLandmarkerResult(&cpp_result);
+
+  FaceLandmarkerResult c_result;
+  CppConvertToFaceLandmarkerResult(cpp_result, &c_result);
+
+  // Verify conversion of face_landmarks
+  EXPECT_EQ(c_result.face_landmarks_count, cpp_result.face_landmarks.size());
+  for (uint32_t i = 0; i < c_result.face_landmarks_count; ++i) {
+    EXPECT_EQ(c_result.face_landmarks[i].landmarks_count,
+              cpp_result.face_landmarks[i].landmarks.size());
+    for (uint32_t j = 0; j < c_result.face_landmarks[i].landmarks_count; ++j) {
+      const auto& cpp_landmark = cpp_result.face_landmarks[i].landmarks[j];
+      EXPECT_FLOAT_EQ(c_result.face_landmarks[i].landmarks[j].x,
+                      cpp_landmark.x);
+      EXPECT_FLOAT_EQ(c_result.face_landmarks[i].landmarks[j].y,
+                      cpp_landmark.y);
+      EXPECT_FLOAT_EQ(c_result.face_landmarks[i].landmarks[j].z,
+                      cpp_landmark.z);
+    }
+  }
+
+  // Verify conversion of face_blendshapes
+  EXPECT_EQ(c_result.face_blendshapes_count,
+            cpp_result.face_blendshapes.value().size());
+  for (uint32_t i = 0; i < c_result.face_blendshapes_count; ++i) {
+    const auto& cpp_face_blendshapes = cpp_result.face_blendshapes.value();
+    EXPECT_EQ(c_result.face_blendshapes[i].categories_count,
+              cpp_face_blendshapes[i].categories.size());
+    for (uint32_t j = 0; j < c_result.face_blendshapes[i].categories_count;
+         ++j) {
+      const auto& cpp_category = cpp_face_blendshapes[i].categories[j];
+      EXPECT_EQ(c_result.face_blendshapes[i].categories[j].index,
+                cpp_category.index);
+      EXPECT_FLOAT_EQ(c_result.face_blendshapes[i].categories[j].score,
+                      cpp_category.score);
+      EXPECT_EQ(
+          std::string(c_result.face_blendshapes[i].categories[j].category_name),
+          cpp_category.category_name);
+    }
+  }
+
+  // Verify conversion of facial_transformation_matrixes
+  EXPECT_EQ(c_result.facial_transformation_matrixes_count,
+            cpp_result.facial_transformation_matrixes.value().size());
+  for (uint32_t i = 0; i < c_result.facial_transformation_matrixes_count; ++i) {
+    const auto& cpp_facial_transformation_matrixes =
+        cpp_result.facial_transformation_matrixes.value();
+    // Assuming Matrix struct contains data array and dimensions
+    const auto& cpp_matrix = cpp_facial_transformation_matrixes[i];
+    EXPECT_EQ(c_result.facial_transformation_matrixes[i].rows,
+              cpp_matrix.rows());
+    EXPECT_EQ(c_result.facial_transformation_matrixes[i].cols,
+              cpp_matrix.cols());
+    // Check each element of the matrix
+    for (int32_t row = 0; row < cpp_matrix.rows(); ++row) {
+      for (int32_t col = 0; col < cpp_matrix.cols(); ++col) {
+        size_t index = col * cpp_matrix.rows() + row;  // Column-major index
+        EXPECT_FLOAT_EQ(c_result.facial_transformation_matrixes[i].data[index],
+                        cpp_matrix(row, col));
+      }
+    }
+  }
+
+  CppCloseFaceLandmarkerResult(&c_result);
+}
+
+TEST(FaceLandmarkerResultConverterTest, FreesMemory) {
+  ::mediapipe::tasks::vision::face_landmarker::FaceLandmarkerResult cpp_result;
+  InitFaceLandmarkerResult(&cpp_result);
+
+  FaceLandmarkerResult c_result;
+  CppConvertToFaceLandmarkerResult(cpp_result, &c_result);
+
+  EXPECT_NE(c_result.face_blendshapes, nullptr);
+  EXPECT_NE(c_result.face_landmarks, nullptr);
+  EXPECT_NE(c_result.facial_transformation_matrixes, nullptr);
+
+  CppCloseFaceLandmarkerResult(&c_result);
+
+  EXPECT_EQ(c_result.face_blendshapes, nullptr);
+  EXPECT_EQ(c_result.face_landmarks, nullptr);
+  EXPECT_EQ(c_result.facial_transformation_matrixes, nullptr);
+}
+
+}  // namespace mediapipe::tasks::c::components::containers

mediapipe/tasks/c/vision/face_landmarker/face_landmarker_test.cc

@@ -0,0 +1,292 @@
+/* Copyright 2023 The MediaPipe Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker.h"
+
+#include <cstdint>
+#include <cstdio>
+#include <cstdlib>
+#include <string>
+
+#include "absl/flags/flag.h"
+#include "absl/strings/string_view.h"
+#include "mediapipe/framework/deps/file_path.h"
+#include "mediapipe/framework/formats/image.h"
+#include "mediapipe/framework/port/gmock.h"
+#include "mediapipe/framework/port/gtest.h"
+#include "mediapipe/tasks/c/components/containers/landmark.h"
+#include "mediapipe/tasks/c/vision/core/common.h"
+#include "mediapipe/tasks/c/vision/face_landmarker/face_landmarker_result.h"
+#include "mediapipe/tasks/cc/vision/utils/image_utils.h"
+
+namespace {
+
+using ::mediapipe::file::JoinPath;
+using ::mediapipe::tasks::vision::DecodeImageFromFile;
+using testing::HasSubstr;
+
+constexpr char kTestDataDirectory[] = "/mediapipe/tasks/testdata/vision/";
+constexpr char kModelName[] = "face_landmarker_v2_with_blendshapes.task";
+constexpr char kImageFile[] = "portrait.jpg";
+constexpr float kLandmarksPrecision = 0.03;
+constexpr float kBlendshapesPrecision = 0.12;
+constexpr float kFacialTransformationMatrixPrecision = 0.05;
+constexpr int kIterations = 100;
+
+std::string GetFullPath(absl::string_view file_name) {
+  return JoinPath("./", kTestDataDirectory, file_name);
+}
+
+void AssertHandLandmarkerResult(const FaceLandmarkerResult* result,
+                                const float blendshapes_precision,
+                                const float landmark_precision,
+                                const float matrix_precison) {
+  // Expects to have the same number of faces detected.
+  EXPECT_EQ(result->face_blendshapes_count, 1);
+
+  // Actual blendshapes matches expected blendshapes.
+  EXPECT_EQ(
+      std::string{result->face_blendshapes[0].categories[0].category_name},
+      "_neutral");
+  EXPECT_NEAR(result->face_blendshapes[0].categories[0].score, 0.0f,
+              blendshapes_precision);
+
+  // Actual landmarks match expected landmarks.
+  EXPECT_NEAR(result->face_landmarks[0].landmarks[0].x, 0.4977f,
+              landmark_precision);
+  EXPECT_NEAR(result->face_landmarks[0].landmarks[0].y, 0.2485f,
+              landmark_precision);
+  EXPECT_NEAR(result->face_landmarks[0].landmarks[0].z, -0.0305f,
+              landmark_precision);
+
+  // Expects to have at least one facial transformation matrix.
+  EXPECT_GE(result->facial_transformation_matrixes_count, 1);
+
+  // Actual matrix matches expected matrix.
+  // Assuming the expected matrix is 2x2 for demonstration.
+  const float expected_matrix[4] = {0.9991f, 0.0166f, -0.0374f, 0.0f};
+  for (int i = 0; i < 4; ++i) {
+    printf(">> %f <<", result->facial_transformation_matrixes[0].data[i]);
+    EXPECT_NEAR(result->facial_transformation_matrixes[0].data[i],
+                expected_matrix[i], matrix_precison);
+  }
+}
+
+TEST(FaceLandmarkerTest, ImageModeTest) {
+  const auto image = DecodeImageFromFile(GetFullPath(kImageFile));
+  ASSERT_TRUE(image.ok());
+
+  const std::string model_path = GetFullPath(kModelName);
+  FaceLandmarkerOptions options = {
+      /* base_options= */ {/* model_asset_buffer= */ nullptr,
+                           /* model_asset_buffer_count= */ 0,
+                           /* model_asset_path= */ model_path.c_str()},
+      /* running_mode= */ RunningMode::IMAGE,
+      /* num_faces= */ 1,
+      /* min_face_detection_confidence= */ 0.5,
+      /* min_face_presence_confidence= */ 0.5,
+      /* min_tracking_confidence= */ 0.5,
+      /* output_face_blendshapes = */ true,
+      /* output_facial_transformation_matrixes = */ true,
+  };
+
+  void* landmarker = face_landmarker_create(&options, /* error_msg */ nullptr);
+  EXPECT_NE(landmarker, nullptr);
+
+  const auto& image_frame = image->GetImageFrameSharedPtr();
+  const MpImage mp_image = {
+      .type = MpImage::IMAGE_FRAME,
+      .image_frame = {.format = static_cast<ImageFormat>(image_frame->Format()),
+                      .image_buffer = image_frame->PixelData(),
+                      .width = image_frame->Width(),
+                      .height = image_frame->Height()}};
+
+  FaceLandmarkerResult result;
+  face_landmarker_detect_image(landmarker, mp_image, &result,
+                               /* error_msg */ nullptr);
+  AssertHandLandmarkerResult(&result, kBlendshapesPrecision,
+                             kLandmarksPrecision,
+                             kFacialTransformationMatrixPrecision);
+  face_landmarker_close_result(&result);
+  face_landmarker_close(landmarker, /* error_msg */ nullptr);
+}
+
+TEST(FaceLandmarkerTest, VideoModeTest) {
+  const auto image = DecodeImageFromFile(GetFullPath(kImageFile));
+  ASSERT_TRUE(image.ok());
+
+  const std::string model_path = GetFullPath(kModelName);
+  FaceLandmarkerOptions options = {
+      /* base_options= */ {/* model_asset_buffer= */ nullptr,
+                           /* model_asset_buffer_count= */ 0,
+                           /* model_asset_path= */ model_path.c_str()},
+      /* running_mode= */ RunningMode::VIDEO,
+      /* num_faces= */ 1,
+      /* min_face_detection_confidence= */ 0.5,
+      /* min_face_presence_confidence= */ 0.5,
+      /* min_tracking_confidence= */ 0.5,
+      /* output_face_blendshapes = */ true,
+      /* output_facial_transformation_matrixes = */ true,
+  };
+
+  void* landmarker = face_landmarker_create(&options,
+                                            /* error_msg */ nullptr);
+  EXPECT_NE(landmarker, nullptr);
+
+  const auto& image_frame = image->GetImageFrameSharedPtr();
+  const MpImage mp_image = {
+      .type = MpImage::IMAGE_FRAME,
+      .image_frame = {.format = static_cast<ImageFormat>(image_frame->Format()),
+                      .image_buffer = image_frame->PixelData(),
+                      .width = image_frame->Width(),
+                      .height = image_frame->Height()}};
+
+  for (int i = 0; i < kIterations; ++i) {
+    FaceLandmarkerResult result;
+    face_landmarker_detect_for_video(landmarker, mp_image, i, &result,
+                                     /* error_msg */ nullptr);
+
+    AssertHandLandmarkerResult(&result, kBlendshapesPrecision,
+                               kLandmarksPrecision,
+                               kFacialTransformationMatrixPrecision);
+    face_landmarker_close_result(&result);
+  }
+  face_landmarker_close(landmarker, /* error_msg */ nullptr);
+}
+
+// A structure to support LiveStreamModeTest below. This structure holds a
+// static method `Fn` for a callback function of C API. A `static` qualifier
+// allows to take an address of the method to follow API style. Another static
+// struct member is `last_timestamp` that is used to verify that current
+// timestamp is greater than the previous one.
+struct LiveStreamModeCallback {
+  static int64_t last_timestamp;
+  static void Fn(const FaceLandmarkerResult* landmarker_result,
+                 const MpImage& image, int64_t timestamp, char* error_msg) {
+    ASSERT_NE(landmarker_result, nullptr);
+    ASSERT_EQ(error_msg, nullptr);
+    AssertHandLandmarkerResult(landmarker_result, kBlendshapesPrecision,
+                               kLandmarksPrecision,
+                               kFacialTransformationMatrixPrecision);
+    EXPECT_GT(image.image_frame.width, 0);
+    EXPECT_GT(image.image_frame.height, 0);
+    EXPECT_GT(timestamp, last_timestamp);
+    ++last_timestamp;
+  }
+};
+int64_t LiveStreamModeCallback::last_timestamp = -1;
+
+TEST(FaceLandmarkerTest, LiveStreamModeTest) {
+  const auto image = DecodeImageFromFile(GetFullPath(kImageFile));
+  ASSERT_TRUE(image.ok());
+
+  const std::string model_path = GetFullPath(kModelName);
+
+  FaceLandmarkerOptions options = {
+      /* base_options= */ {/* model_asset_buffer= */ nullptr,
+                           /* model_asset_buffer_count= */ 0,
+                           /* model_asset_path= */ model_path.c_str()},
+      /* running_mode= */ RunningMode::LIVE_STREAM,
+      /* num_faces= */ 1,
+      /* min_face_detection_confidence= */ 0.5,
+      /* min_face_presence_confidence= */ 0.5,
+      /* min_tracking_confidence= */ 0.5,
+      /* output_face_blendshapes = */ true,
+      /* output_facial_transformation_matrixes = */ true,
+      /* result_callback= */ LiveStreamModeCallback::Fn,
+  };
+
+  void* landmarker = face_landmarker_create(&options, /* error_msg */
+                                            nullptr);
+  EXPECT_NE(landmarker, nullptr);
+
+  const auto& image_frame = image->GetImageFrameSharedPtr();
+  const MpImage mp_image = {
+      .type = MpImage::IMAGE_FRAME,
+      .image_frame = {.format = static_cast<ImageFormat>(image_frame->Format()),
+                      .image_buffer = image_frame->PixelData(),
+                      .width = image_frame->Width(),
+                      .height = image_frame->Height()}};
+
+  for (int i = 0; i < kIterations; ++i) {
+    EXPECT_GE(face_landmarker_detect_async(landmarker, mp_image, i,
+                                           /* error_msg */ nullptr),
+              0);
+  }
+  face_landmarker_close(landmarker, /* error_msg */ nullptr);
+
+  // Due to the flow limiter, the total of outputs might be smaller than the
+  // number of iterations.
+  EXPECT_LE(LiveStreamModeCallback::last_timestamp, kIterations);
+  EXPECT_GT(LiveStreamModeCallback::last_timestamp, 0);
+}
+
+TEST(FaceLandmarkerTest, InvalidArgumentHandling) {
+  // It is an error to set neither the asset buffer nor the path.
+  FaceLandmarkerOptions options = {
+      /* base_options= */ {/* model_asset_buffer= */ nullptr,
+                           /* model_asset_buffer_count= */ 0,
+                           /* model_asset_path= */ nullptr},
+      /* running_mode= */ RunningMode::IMAGE,
+      /* num_faces= */ 1,
+      /* min_face_detection_confidence= */ 0.5,
+      /* min_face_presence_confidence= */ 0.5,
+      /* min_tracking_confidence= */ 0.5,
+      /* output_face_blendshapes = */ true,
+      /* output_facial_transformation_matrixes = */ true,
+  };
+
+  char* error_msg;
+  void* landmarker = face_landmarker_create(&options, &error_msg);
+  EXPECT_EQ(landmarker, nullptr);
+
+  EXPECT_THAT(
+      error_msg,
+      HasSubstr("INVALID_ARGUMENT: BLENDSHAPES Tag and blendshapes model must "
+                "be both set. Get BLENDSHAPES is set: true, blendshapes model "
+                "is set: false [MediaPipeTasksStatus='601']"));
+
+  free(error_msg);
+}
+
+TEST(FaceLandmarkerTest, FailedRecognitionHandling) {
+  const std::string model_path = GetFullPath(kModelName);
+  FaceLandmarkerOptions options = {
+      /* base_options= */ {/* model_asset_buffer= */ nullptr,
+                           /* model_asset_buffer_count= */ 0,
+                           /* model_asset_path= */ model_path.c_str()},
+      /* running_mode= */ RunningMode::IMAGE,
+      /* num_faces= */ 1,
+      /* min_face_detection_confidence= */ 0.5,
+      /* min_face_presence_confidence= */ 0.5,
+      /* min_tracking_confidence= */ 0.5,
+      /* output_face_blendshapes = */ true,
+      /* output_facial_transformation_matrixes = */ true,
+  };
+
+  void* landmarker = face_landmarker_create(&options, /* error_msg */
+                                            nullptr);
+  EXPECT_NE(landmarker, nullptr);
+
+  const MpImage mp_image = {.type = MpImage::GPU_BUFFER, .gpu_buffer = {}};
+  FaceLandmarkerResult result;
+  char* error_msg;
+  face_landmarker_detect_image(landmarker, mp_image, &result, &error_msg);
+  EXPECT_THAT(error_msg, HasSubstr("GPU Buffer not supported yet"));
+  free(error_msg);
+  face_landmarker_close(landmarker, /* error_msg */ nullptr);
+}
+
+}  // namespace

mediapipe/tasks/c/vision/hand_landmarker/hand_landmarker.h

@@ -68,7 +68,7 @@ struct HandLandmarkerOptions {
   // true for the lifetime of the callback function.
   //
   // A caller is responsible for closing hand landmarker result.
-  typedef void (*result_callback_fn)(HandLandmarkerResult* result,
+  typedef void (*result_callback_fn)(const HandLandmarkerResult* result,
                                      const MpImage& image, int64_t timestamp_ms,
                                      char* error_msg);
   result_callback_fn result_callback;

mediapipe/tasks/c/vision/hand_landmarker/hand_landmarker_test.cc

@@ -47,7 +47,7 @@ std::string GetFullPath(absl::string_view file_name) {
   return JoinPath("./", kTestDataDirectory, file_name);
 }
 
-void MatchesHandLandmarkerResult(HandLandmarkerResult* result,
+void AssertHandLandmarkerResult(const HandLandmarkerResult* result,
                                 const float score_precision,
                                 const float landmark_precision) {
   // Expects to have the same number of hands detected.
@@ -104,7 +104,7 @@ TEST(HandLandmarkerTest, ImageModeTest) {
   HandLandmarkerResult result;
   hand_landmarker_detect_image(landmarker, mp_image, &result,
                                /* error_msg */ nullptr);
-  MatchesHandLandmarkerResult(&result, kScorePrecision, kLandmarkPrecision);
+  AssertHandLandmarkerResult(&result, kScorePrecision, kLandmarkPrecision);
   hand_landmarker_close_result(&result);
   hand_landmarker_close(landmarker, /* error_msg */ nullptr);
 }
@@ -141,7 +141,7 @@ TEST(HandLandmarkerTest, VideoModeTest) {
     hand_landmarker_detect_for_video(landmarker, mp_image, i, &result,
                                      /* error_msg */ nullptr);
 
-    MatchesHandLandmarkerResult(&result, kScorePrecision, kLandmarkPrecision);
+    AssertHandLandmarkerResult(&result, kScorePrecision, kLandmarkPrecision);
     hand_landmarker_close_result(&result);
   }
   hand_landmarker_close(landmarker, /* error_msg */ nullptr);
@@ -154,11 +154,11 @@ TEST(HandLandmarkerTest, VideoModeTest) {
 // timestamp is greater than the previous one.
 struct LiveStreamModeCallback {
   static int64_t last_timestamp;
-  static void Fn(HandLandmarkerResult* landmarker_result, const MpImage& image,
-                 int64_t timestamp, char* error_msg) {
+  static void Fn(const HandLandmarkerResult* landmarker_result,
+                 const MpImage& image, int64_t timestamp, char* error_msg) {
     ASSERT_NE(landmarker_result, nullptr);
     ASSERT_EQ(error_msg, nullptr);
-    MatchesHandLandmarkerResult(landmarker_result, kScorePrecision,
+    AssertHandLandmarkerResult(landmarker_result, kScorePrecision,
                                kLandmarkPrecision);
     EXPECT_GT(image.image_frame.width, 0);
     EXPECT_GT(image.image_frame.height, 0);
@@ -183,7 +183,7 @@ TEST(HandLandmarkerTest, LiveStreamModeTest) {
       /* min_hand_detection_confidence= */ 0.5,
       /* min_hand_presence_confidence= */ 0.5,
       /* min_tracking_confidence= */ 0.5,
-      /* result_callback= */ LiveStreamModeCallback::Fn,
+      /* result_callback_fn= */ LiveStreamModeCallback::Fn,
   };
 
   void* landmarker = hand_landmarker_create(&options, /* error_msg */ nullptr);

mediapipe/tasks/cc/vision/image_generator/diffuser/stable_diffusion_iterate_calculator.cc

@@ -121,6 +121,8 @@ class StableDiffusionIterateCalculator : public Node {
     if (handle_) dlclose(handle_);
   }
 
+  static absl::Status UpdateContract(CalculatorContract* cc);
+
   absl::Status Open(CalculatorContext* cc) override;
   absl::Status Process(CalculatorContext* cc) override;
 
@@ -188,6 +190,11 @@ class StableDiffusionIterateCalculator : public Node {
   bool emit_empty_packet_;
 };
 
+absl::Status StableDiffusionIterateCalculator::UpdateContract(
+    CalculatorContract* cc) {
+  return absl::OkStatus();
+}
+
 absl::Status StableDiffusionIterateCalculator::Open(CalculatorContext* cc) {
   StableDiffusionIterateCalculatorOptions options;
   if (kOptionsIn(cc).IsEmpty()) {
@@ -205,7 +212,11 @@ absl::Status StableDiffusionIterateCalculator::Open(CalculatorContext* cc) {
   if (options.file_folder().empty()) {
     std::strcpy(config.model_dir, "bins/");  // NOLINT
   } else {
-    std::strcpy(config.model_dir, options.file_folder().c_str());  // NOLINT
+    std::string file_folder = options.file_folder();
+    if (!file_folder.empty() && file_folder.back() != '/') {
+      file_folder.push_back('/');
+    }
+    std::strcpy(config.model_dir, file_folder.c_str());  // NOLINT
   }
   MP_RETURN_IF_ERROR(mediapipe::file::Exists(config.model_dir))
       << config.model_dir;

mediapipe/tasks/ios/BUILD

@@ -59,6 +59,7 @@ CALCULATORS_AND_GRAPHS = [
     "//mediapipe/tasks/cc/vision/image_classifier:image_classifier_graph",
     "//mediapipe/tasks/cc/vision/image_segmenter:image_segmenter_graph",
     "//mediapipe/tasks/cc/vision/object_detector:object_detector_graph",
+    "//mediapipe/tasks/cc/vision/pose_landmarker:pose_landmarker_graph",
 ]
 
 strip_api_include_path_prefix(
@@ -106,6 +107,9 @@ strip_api_include_path_prefix(
         "//mediapipe/tasks/ios/vision/object_detector:sources/MPPObjectDetector.h",
         "//mediapipe/tasks/ios/vision/object_detector:sources/MPPObjectDetectorOptions.h",
         "//mediapipe/tasks/ios/vision/object_detector:sources/MPPObjectDetectorResult.h",
+        "//mediapipe/tasks/ios/vision/pose_landmarker:sources/MPPPoseLandmarker.h",
+        "//mediapipe/tasks/ios/vision/pose_landmarker:sources/MPPPoseLandmarkerOptions.h",
+        "//mediapipe/tasks/ios/vision/pose_landmarker:sources/MPPPoseLandmarkerResult.h",
     ],
 )
 
@@ -206,6 +210,9 @@ apple_static_xcframework(
         ":MPPObjectDetector.h",
         ":MPPObjectDetectorOptions.h",
         ":MPPObjectDetectorResult.h",
+        ":MPPPoseLandmarker.h",
+        ":MPPPoseLandmarkerOptions.h",
+        ":MPPPoseLandmarkerResult.h",
     ],
     deps = [
         "//mediapipe/tasks/ios/vision/face_detector:MPPFaceDetector",
@@ -215,6 +222,7 @@ apple_static_xcframework(
         "//mediapipe/tasks/ios/vision/image_classifier:MPPImageClassifier",
         "//mediapipe/tasks/ios/vision/image_segmenter:MPPImageSegmenter",
         "//mediapipe/tasks/ios/vision/object_detector:MPPObjectDetector",
+        "//mediapipe/tasks/ios/vision/pose_landmarker:MPPPoseLandmarker",
     ],
 )
 

mediapipe/tasks/ios/vision/core/utils/sources/MPPImage+Utils.mm

@@ -30,6 +30,20 @@ namespace {
 using ::mediapipe::ImageFormat;
 using ::mediapipe::ImageFrame;
 
+vImage_Buffer CreateEmptyVImageBufferFromImageFrame(ImageFrame &imageFrame, bool shouldAllocate) {
+  UInt8 *data = shouldAllocate ? new UInt8[imageFrame.Height() * imageFrame.WidthStep()] : nullptr;
+  return {.data = data,
+          .height = static_cast<vImagePixelCount>(imageFrame.Height()),
+          .width = static_cast<vImagePixelCount>(imageFrame.Width()),
+          .rowBytes = static_cast<size_t>(imageFrame.WidthStep())};
+}
+
+vImage_Buffer CreateVImageBufferFromImageFrame(ImageFrame &imageFrame) {
+  vImage_Buffer imageBuffer = CreateEmptyVImageBufferFromImageFrame(imageFrame, false);
+  imageBuffer.data = imageFrame.MutablePixelData();
+  return imageBuffer;
+}
+
 vImage_Buffer allocatedVImageBuffer(vImagePixelCount width, vImagePixelCount height,
                                     size_t rowBytes) {
   UInt8 *data = new UInt8[height * rowBytes];
@@ -40,6 +54,8 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
   delete[] (vImage_Buffer *)buffer;
 }
 
+static void FreeRefConReleaseCallback(void *refCon, const void *baseAddress) { free(refCon); }
+
 }  // namespace
 
 @interface MPPPixelDataUtils : NSObject
@@ -51,6 +67,10 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
                                      pixelBufferFormat:(OSType)pixelBufferFormatType
                                                  error:(NSError **)error;
 
++ (UInt8 *)pixelDataFromImageFrame:(ImageFrame &)imageFrame
+                        shouldCopy:(BOOL)shouldCopy
+                             error:(NSError **)error;
+
 @end
 
 @interface MPPCVPixelBufferUtils : NSObject
@@ -58,6 +78,24 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
 + (std::unique_ptr<ImageFrame>)imageFrameFromCVPixelBuffer:(CVPixelBufferRef)pixelBuffer
                                                      error:(NSError **)error;
 
+// This method is used to create CVPixelBuffer from output images of tasks like `FaceStylizer` only
+// when the input `MPImage` source type is `pixelBuffer`.
+// Always copies the pixel data of the image frame to the created `CVPixelBuffer`.
+//
+// The only possible 32 RGBA pixel format of input `CVPixelBuffer` is `kCVPixelFormatType_32BGRA`.
+// But Mediapipe does not support inference on images of format `BGRA`. Hence the channels of the
+// underlying pixel data of `CVPixelBuffer` are permuted to the supported RGBA format before passing
+// them to the task for inference. The pixel format of the output images of any MediaPipe task will
+// be the same as the pixel format of the input image. (RGBA in this case).
+//
+// Since creation of `CVPixelBuffer` from the output image pixels with a format of
+// `kCVPixelFormatType_32RGBA` is not possible, the channels of the output C++ image `RGBA` have to
+// be permuted to the format `BGRA`. When the pixels are copied to create `CVPixelBuffer` this does
+// not pose a challenge.
+//
+// TODO: Investigate if permuting channels of output `mediapipe::Image` in place is possible for
+// creating `CVPixelBuffer`s without copying the underlying pixels.
++ (CVPixelBufferRef)cvPixelBufferFromImageFrame:(ImageFrame &)imageFrame error:(NSError **)error;
 @end
 
 @interface MPPCGImageUtils : NSObject
@@ -99,6 +137,9 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
 
   // Convert the raw pixel data to RGBA format and un-premultiply the alpha from the R, G, B values
   // since MediaPipe C++ APIs only accept un pre-multiplied channels.
+  //
+  // This method is commonly used for `MPImage`s of all source types. Hence supporting BGRA and RGBA
+  // formats. Only `pixelBuffer` source type is restricted to `BGRA` format.
   switch (pixelBufferFormatType) {
     case kCVPixelFormatType_32RGBA: {
       destBuffer = allocatedVImageBuffer((vImagePixelCount)width, (vImagePixelCount)height,
@@ -107,6 +148,8 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
       break;
     }
     case kCVPixelFormatType_32BGRA: {
+      // Permute channels to `RGBA` since MediaPipe tasks don't support inference on images of
+      // format `BGRA`.
       const uint8_t permute_map[4] = {2, 1, 0, 3};
       destBuffer = allocatedVImageBuffer((vImagePixelCount)width, (vImagePixelCount)height,
                                          destinationBytesPerRow);
@@ -120,8 +163,7 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
     default: {
       [MPPCommonUtils createCustomError:error
                                withCode:MPPTasksErrorCodeInvalidArgumentError
-                            description:@"Invalid source pixel buffer format. Expecting one of "
-                                        @"kCVPixelFormatType_32RGBA, kCVPixelFormatType_32BGRA"];
+                            description:@"Some internal error occured."];
       return nullptr;
     }
   }
@@ -139,6 +181,46 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
                                       static_cast<uint8 *>(destBuffer.data));
 }
 
++ (UInt8 *)pixelDataFromImageFrame:(ImageFrame &)imageFrame
+                        shouldCopy:(BOOL)shouldCopy
+                             error:(NSError **)error {
+  vImage_Buffer sourceBuffer = CreateVImageBufferFromImageFrame(imageFrame);
+
+  // Pre-multiply the raw pixels from a `mediapipe::Image` before creating a `CGImage` to ensure
+  // that pixels are displayed correctly irrespective of their alpha values.
+  vImage_Error premultiplyError;
+  vImage_Buffer destinationBuffer;
+
+  switch (imageFrame.Format()) {
+    case ImageFormat::SRGBA: {
+      destinationBuffer =
+          shouldCopy ? CreateEmptyVImageBufferFromImageFrame(imageFrame, true) : sourceBuffer;
+      premultiplyError =
+          vImagePremultiplyData_RGBA8888(&sourceBuffer, &destinationBuffer, kvImageNoFlags);
+      break;
+    }
+    default: {
+      [MPPCommonUtils createCustomError:error
+                               withCode:MPPTasksErrorCodeInternalError
+                            description:@"An error occured while processing the output image "
+                                        @"pixels of the vision task."];
+      return nullptr;
+    }
+  }
+
+  if (premultiplyError != kvImageNoError) {
+    [MPPCommonUtils
+        createCustomError:error
+                 withCode:MPPTasksErrorCodeInternalError
+              description:
+                  @"An error occured while processing the output image pixels of the vision task."];
+
+    return nullptr;
+  }
+
+  return (UInt8 *)destinationBuffer.data;
+}
+
 @end
 
 @implementation MPPCVPixelBufferUtils
@@ -149,7 +231,8 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
   std::unique_ptr<ImageFrame> imageFrame = nullptr;
 
   switch (pixelBufferFormat) {
-    case kCVPixelFormatType_32RGBA:
+    // Core Video only supports pixel data of order BGRA for 32 bit RGBA images.
+    // Thus other formats like `kCVPixelFormatType_32BGRA` don't need to be accounted for.
     case kCVPixelFormatType_32BGRA: {
       CVPixelBufferLockBaseAddress(pixelBuffer, 0);
       imageFrame = [MPPPixelDataUtils
@@ -165,15 +248,58 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
     default: {
       [MPPCommonUtils createCustomError:error
                                withCode:MPPTasksErrorCodeInvalidArgumentError
-                            description:@"Unsupported pixel format for CVPixelBuffer. Supported "
-                                        @"pixel format types are kCVPixelFormatType_32BGRA and "
-                                        @"kCVPixelFormatType_32RGBA"];
+                            description:@"Unsupported pixel format for CVPixelBuffer. Expected "
+                                        @"kCVPixelFormatType_32BGRA"];
     }
   }
 
   return imageFrame;
 }
 
++ (CVPixelBufferRef)cvPixelBufferFromImageFrame:(ImageFrame &)imageFrame error:(NSError **)error {
+  if (imageFrame.Format() != ImageFormat::SRGBA) {
+    [MPPCommonUtils createCustomError:error
+                             withCode:MPPTasksErrorCodeInternalError
+                          description:@"An error occured while creating a CVPixelBuffer from the "
+                                      @"output image of the vision task."];
+    return nullptr;
+  }
+
+  UInt8 *pixelData = [MPPPixelDataUtils pixelDataFromImageFrame:imageFrame
+                                                     shouldCopy:YES
+                                                          error:error];
+
+  if (!pixelData) {
+    return nullptr;
+  }
+
+  const uint8_t permute_map[4] = {2, 1, 0, 3};
+  vImage_Buffer sourceBuffer = CreateEmptyVImageBufferFromImageFrame(imageFrame, NO);
+  sourceBuffer.data = pixelData;
+
+  if (vImagePermuteChannels_ARGB8888(&sourceBuffer, &sourceBuffer, permute_map, kvImageNoFlags) ==
+      kvImageNoError) {
+    CVPixelBufferRef outputBuffer;
+
+    OSType pixelBufferFormatType = kCVPixelFormatType_32BGRA;
+
+    // Since data is copied, pass in a release callback that will be invoked when the pixel buffer
+    // is destroyed.
+    if (CVPixelBufferCreateWithBytes(kCFAllocatorDefault, imageFrame.Width(), imageFrame.Height(),
+                                     pixelBufferFormatType, pixelData, imageFrame.WidthStep(),
+                                     FreeRefConReleaseCallback, pixelData, nullptr,
+                                     &outputBuffer) == kCVReturnSuccess) {
+      return outputBuffer;
+    }
+  }
+
+  [MPPCommonUtils createCustomError:error
+                           withCode:MPPTasksErrorCodeInternalError
+                        description:@"An error occured while creating a CVPixelBuffer from the "
+                                    @"output image of the vision task."];
+  return nullptr;
+}
+
 @end
 
 @implementation MPPCGImageUtils
@@ -232,7 +358,14 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
   CGBitmapInfo bitmapInfo = kCGImageAlphaNoneSkipLast | kCGBitmapByteOrderDefault;
 
   ImageFrame *internalImageFrame = imageFrame.get();
-  size_t channelCount = 4;
+
+  UInt8 *pixelData = [MPPPixelDataUtils pixelDataFromImageFrame:*internalImageFrame
+                                                     shouldCopy:shouldCopyPixelData
+                                                          error:error];
+
+  if (!pixelData) {
+    return nullptr;
+  }
 
   switch (internalImageFrame->Format()) {
     case ImageFormat::SRGBA: {
@@ -242,56 +375,41 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
     default:
       [MPPCommonUtils createCustomError:error
                                withCode:MPPTasksErrorCodeInternalError
-                            description:@"An internal error occured."];
+                            description:@"An error occured while creating a CGImage from the "
+                                        @"output image of the vision task."];
       return nullptr;
   }
 
-  size_t bitsPerComponent = 8;
-
-  vImage_Buffer sourceBuffer = {
-      .data = (void *)internalImageFrame->MutablePixelData(),
-      .height = static_cast<vImagePixelCount>(internalImageFrame->Height()),
-      .width = static_cast<vImagePixelCount>(internalImageFrame->Width()),
-      .rowBytes = static_cast<size_t>(internalImageFrame->WidthStep())};
-
-  vImage_Buffer destBuffer;
-
   CGDataProviderReleaseDataCallback callback = nullptr;
 
-  if (shouldCopyPixelData) {
-    destBuffer = allocatedVImageBuffer(static_cast<vImagePixelCount>(internalImageFrame->Width()),
-                                       static_cast<vImagePixelCount>(internalImageFrame->Height()),
-                                       static_cast<size_t>(internalImageFrame->WidthStep()));
-    callback = FreeDataProviderReleaseCallback;
-  } else {
-    destBuffer = sourceBuffer;
-  }
-
-  // Pre-multiply the raw pixels from a `mediapipe::Image` before creating a `CGImage` to ensure
-  // that pixels are displayed correctly irrespective of their alpha values.
-  vImage_Error premultiplyError =
-      vImagePremultiplyData_RGBA8888(&sourceBuffer, &destBuffer, kvImageNoFlags);
-
-  if (premultiplyError != kvImageNoError) {
-    [MPPCommonUtils createCustomError:error
-                             withCode:MPPTasksErrorCodeInternalError
-                          description:@"An internal error occured."];
-
-    return nullptr;
-  }
-
   CGDataProviderRef provider = CGDataProviderCreateWithData(
-      destBuffer.data, destBuffer.data,
-      internalImageFrame->WidthStep() * internalImageFrame->Height(), callback);
+      pixelData, pixelData, internalImageFrame->WidthStep() * internalImageFrame->Height(),
+      callback);
+
   CGColorSpaceRef colorSpace = CGColorSpaceCreateDeviceRGB();
-  CGImageRef cgImageRef =
+
+  CGImageRef cgImageRef = nullptr;
+
+  if (provider && colorSpace) {
+    size_t bitsPerComponent = 8;
+    size_t channelCount = 4;
+    cgImageRef =
         CGImageCreate(internalImageFrame->Width(), internalImageFrame->Height(), bitsPerComponent,
                       bitsPerComponent * channelCount, internalImageFrame->WidthStep(), colorSpace,
                       bitmapInfo, provider, nullptr, YES, kCGRenderingIntentDefault);
+  }
 
+  // Can safely pass `NULL` to these functions according to iOS docs.
   CGDataProviderRelease(provider);
   CGColorSpaceRelease(colorSpace);
 
+  if (!cgImageRef) {
+    [MPPCommonUtils createCustomError:error
+                             withCode:MPPTasksErrorCodeInternalError
+                          description:@"An error occured while converting the output image of the "
+                                      @"vision task to a CGImage."];
+  }
+
   return cgImageRef;
 }
 
@@ -347,8 +465,30 @@ static void FreeDataProviderReleaseCallback(void *buffer, const void *data, size
 
       return [self initWithUIImage:image orientation:sourceImage.orientation error:nil];
     }
+    case MPPImageSourceTypePixelBuffer: {
+      if (!shouldCopyPixelData) {
+        // TODO: Investigate possibility of permuting channels of `mediapipe::Image` returned by
+        // vision tasks in place to ensure that we can support creating `CVPixelBuffer`s without
+        // copying the pixel data.
+        [MPPCommonUtils
+            createCustomError:error
+                     withCode:MPPTasksErrorCodeInvalidArgumentError
+                  description:
+                      @"When the source type is pixel buffer, you cannot request uncopied data"];
+        return nil;
+      }
+
+      CVPixelBufferRef pixelBuffer =
+          [MPPCVPixelBufferUtils cvPixelBufferFromImageFrame:*(image.GetImageFrameSharedPtr())
+                                                       error:error];
+      MPPImage *image = [self initWithPixelBuffer:pixelBuffer
+                                      orientation:sourceImage.orientation
+                                            error:nil];
+      CVPixelBufferRelease(pixelBuffer);
+      return image;
+    }
     default:
-      // TODO Implement Other Source Types.
+      // TODO Implement CMSampleBuffer.
       return nil;
   }
 }

mediapipe/tasks/ios/vision/image_classifier/sources/MPPImageClassifier.h

@@ -82,10 +82,8 @@ NS_SWIFT_NAME(ImageClassifier)
  * `.image`.
  *
  * This method supports classification of RGBA images. If your `MPImage` has a source type
- * ofm`.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must have one of the following
- * pixel format types:
- * 1. kCVPixelFormatType_32BGRA
- * 2. kCVPixelFormatType_32RGBA
+ * of `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must use
+ * `kCVPixelFormatType_32BGRA` as its pixel format.
  *
  * If your `MPImage` has a source type of `.image` ensure that the color space is RGB with an Alpha
  * channel.
@@ -104,11 +102,9 @@ NS_SWIFT_NAME(ImageClassifier)
  * of the provided `MPImage`. Only use this method when the `ImageClassifier` is created with
  * running mode, `.image`.
  *
- * This method supports classification of RGBA images. If your `MPImage` has a source type of
- * `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must have one of the following
- * pixel format types:
- * 1. kCVPixelFormatType_32BGRA
- * 2. kCVPixelFormatType_32RGBA
+ * This method supports classification of RGBA images. If your `MPImage` has a source type
+ * of `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must use
+ * `kCVPixelFormatType_32BGRA` as its pixel format.
  *
  * If your `MPImage` has a source type of `.image` ensure that the color space is RGB with an Alpha
  * channel.
@@ -133,11 +129,9 @@ NS_SWIFT_NAME(ImageClassifier)
  * It's required to provide the video frame's timestamp (in milliseconds). The input timestamps must
  * be monotonically increasing.
  *
- * This method supports classification of RGBA images. If your `MPImage` has a source type of
- * `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must have one of the following
- * pixel format types:
- * 1. kCVPixelFormatType_32BGRA
- * 2. kCVPixelFormatType_32RGBA
+ * This method supports classification of RGBA images. If your `MPImage` has a source type
+ * of `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must use
+ * `kCVPixelFormatType_32BGRA` as its pixel format.
  *
  * If your `MPImage` has a source type of `.image` ensure that the color space is RGB with an Alpha
  * channel.
@@ -161,11 +155,9 @@ NS_SWIFT_NAME(ImageClassifier)
  * It's required to provide the video frame's timestamp (in milliseconds). The input timestamps must
  * be monotonically increasing.
  *
- * This method supports classification of RGBA images. If your `MPImage` has a source type of
- * `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must have one of the following
- * pixel format types:
- * 1. kCVPixelFormatType_32BGRA
- * 2. kCVPixelFormatType_32RGBA
+ * This method supports classification of RGBA images. If your `MPImage` has a source type
+ * of `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must use
+ * `kCVPixelFormatType_32BGRA` as its pixel format.
  *
  * If your `MPImage` has a source type of `.image` ensure that the color space is RGB with an Alpha
  * channel.
@@ -199,11 +191,9 @@ NS_SWIFT_NAME(ImageClassifier)
  * It's required to provide a timestamp (in milliseconds) to indicate when the input image is sent
  * to the image classifier. The input timestamps must be monotonically increasing.
  *
- * This method supports classification of RGBA images. If your `MPImage` has a source type of
- * .pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must have one of the following
- * pixel format types:
- * 1. kCVPixelFormatType_32BGRA
- * 2. kCVPixelFormatType_32RGBA
+ * This method supports classification of RGBA images. If your `MPImage` has a source type
+ * of `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must use
+ * `kCVPixelFormatType_32BGRA` as its pixel format.
  *
  * If the input `MPImage` has a source type of `.image` ensure that the color space is RGB with an
  * Alpha channel.
@@ -238,17 +228,15 @@ NS_SWIFT_NAME(ImageClassifier)
  * It's required to provide a timestamp (in milliseconds) to indicate when the input image is sent
  * to the image classifier. The input timestamps must be monotonically increasing.
  *
- * This method supports classification of RGBA images. If your `MPImage` has a source type of
- * `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must have one of the following
- * pixel format types:
- * 1. kCVPixelFormatType_32BGRA
- * 2. kCVPixelFormatType_32RGBA
+ * This method supports classification of RGBA images. If your `MPImage` has a source type
+ * of `.pixelBuffer` or `.sampleBuffer`, the underlying pixel buffer must use
+ * `kCVPixelFormatType_32BGRA` as its pixel format.
  *
  * If the input `MPImage` has a source type of `.image` ensure that the color space is RGB with an
  * Alpha channel.
  *
  * If this method is used for classifying live camera frames using `AVFoundation`, ensure that you
- * request `AVCaptureVideoDataOutput` to output frames in `kCMPixelFormat_32RGBA` using its
+ * request `AVCaptureVideoDataOutput` to output frames in `kCMPixelFormat_32BGRA` using its
  * `videoSettings` property.
  *
  * @param image A live stream image data of type `MPImage` on which image classification is to be

mediapipe/tasks/ios/vision/pose_landmarker/utils/sources/MPPPoseLandmarkerOptions+Helpers.mm

@@ -36,7 +36,8 @@ using PoseLandmarkerGraphOptionsProto =
       optionsProto->MutableExtension(PoseLandmarkerGraphOptionsProto::ext);
   poseLandmarkerGraphOptions->Clear();
 
-  [self.baseOptions copyToProto:poseLandmarkerGraphOptions->mutable_base_options()];
+  [self.baseOptions copyToProto:poseLandmarkerGraphOptions->mutable_base_options()
+              withUseStreamMode:self.runningMode != MPPRunningModeImage];
   poseLandmarkerGraphOptions->set_min_tracking_confidence(self.minTrackingConfidence);
 
   PoseLandmarksDetectorGraphOptionsProto *poseLandmarksDetectorGraphOptions =

mediapipe/tasks/java/com/google/mediapipe/tasks/vision/holisticlandmarker/HolisticLandmarker.java

@@ -479,7 +479,7 @@ public final class HolisticLandmarker extends BaseVisionTaskApi {
        * Sets minimum confidence score for the face landmark detection to be considered successful.
        * Defaults to 0.5.
        */
-      public abstract Builder setMinFaceLandmarksConfidence(Float value);
+      public abstract Builder setMinFacePresenceConfidence(Float value);
 
       /**
        * The minimum confidence score for the pose detection to be considered successful. Defaults
@@ -497,7 +497,7 @@ public final class HolisticLandmarker extends BaseVisionTaskApi {
        * The minimum confidence score for the pose landmarks detection to be considered successful.
        * Defaults to 0.5.
        */
-      public abstract Builder setMinPoseLandmarksConfidence(Float value);
+      public abstract Builder setMinPosePresenceConfidence(Float value);
 
       /**
        * The minimum confidence score for the hand landmark detection to be considered successful.
@@ -555,13 +555,13 @@ public final class HolisticLandmarker extends BaseVisionTaskApi {
 
     abstract Optional<Float> minFaceSuppressionThreshold();
 
-    abstract Optional<Float> minFaceLandmarksConfidence();
+    abstract Optional<Float> minFacePresenceConfidence();
 
     abstract Optional<Float> minPoseDetectionConfidence();
 
     abstract Optional<Float> minPoseSuppressionThreshold();
 
-    abstract Optional<Float> minPoseLandmarksConfidence();
+    abstract Optional<Float> minPosePresenceConfidence();
 
     abstract Optional<Float> minHandLandmarksConfidence();
 
@@ -578,10 +578,10 @@ public final class HolisticLandmarker extends BaseVisionTaskApi {
           .setRunningMode(RunningMode.IMAGE)
           .setMinFaceDetectionConfidence(DEFAULT_PRESENCE_THRESHOLD)
           .setMinFaceSuppressionThreshold(DEFAULT_SUPPRESION_THRESHOLD)
-          .setMinFaceLandmarksConfidence(DEFAULT_PRESENCE_THRESHOLD)
+          .setMinFacePresenceConfidence(DEFAULT_PRESENCE_THRESHOLD)
           .setMinPoseDetectionConfidence(DEFAULT_PRESENCE_THRESHOLD)
           .setMinPoseSuppressionThreshold(DEFAULT_SUPPRESION_THRESHOLD)
-          .setMinPoseLandmarksConfidence(DEFAULT_PRESENCE_THRESHOLD)
+          .setMinPosePresenceConfidence(DEFAULT_PRESENCE_THRESHOLD)
           .setMinHandLandmarksConfidence(DEFAULT_PRESENCE_THRESHOLD)
           .setOutputFaceBlendshapes(DEFAULT_OUTPUT_FACE_BLENDSHAPES)
           .setOutputPoseSegmentationMasks(DEFAULT_OUTPUT_SEGMENTATION_MASKS);
@@ -616,12 +616,12 @@ public final class HolisticLandmarker extends BaseVisionTaskApi {
       // Configure pose detector options.
       minPoseDetectionConfidence().ifPresent(poseDetectorGraphOptions::setMinDetectionConfidence);
       minPoseSuppressionThreshold().ifPresent(poseDetectorGraphOptions::setMinSuppressionThreshold);
-      minPoseLandmarksConfidence().ifPresent(poseLandmarkerGraphOptions::setMinDetectionConfidence);
+      minPosePresenceConfidence().ifPresent(poseLandmarkerGraphOptions::setMinDetectionConfidence);
 
       // Configure face detector options.
       minFaceDetectionConfidence().ifPresent(faceDetectorGraphOptions::setMinDetectionConfidence);
       minFaceSuppressionThreshold().ifPresent(faceDetectorGraphOptions::setMinSuppressionThreshold);
-      minFaceLandmarksConfidence()
+      minFacePresenceConfidence()
           .ifPresent(faceLandmarksDetectorGraphOptions::setMinDetectionConfidence);
 
       holisticLandmarkerGraphOptions

mediapipe/tasks/python/core/BUILD

@@ -49,5 +49,6 @@ py_library(
         "//mediapipe/calculators/core:flow_limiter_calculator_py_pb2",
         "//mediapipe/framework:calculator_options_py_pb2",
         "//mediapipe/framework:calculator_py_pb2",
+        "@com_google_protobuf//:protobuf_python",
     ],
 )

mediapipe/tasks/python/core/task_info.py

@@ -14,9 +14,8 @@
 """MediaPipe Tasks' task info data class."""
 
 import dataclasses
-
 from typing import Any, List
-
+from google.protobuf import any_pb2
 from mediapipe.calculators.core import flow_limiter_calculator_pb2
 from mediapipe.framework import calculator_options_pb2
 from mediapipe.framework import calculator_pb2
@@ -80,21 +79,34 @@ class TaskInfo:
       raise ValueError(
           '`task_options` doesn`t provide `to_pb2()` method to convert itself to be a protobuf object.'
       )
-    task_subgraph_options = calculator_options_pb2.CalculatorOptions()
+
     task_options_proto = self.task_options.to_pb2()
-    task_subgraph_options.Extensions[task_options_proto.ext].CopyFrom(
-        task_options_proto)
-    if not enable_flow_limiting:
-      return calculator_pb2.CalculatorGraphConfig(
-          node=[
-              calculator_pb2.CalculatorGraphConfig.Node(
+
+    node_config = calculator_pb2.CalculatorGraphConfig.Node(
         calculator=self.task_graph,
         input_stream=self.input_streams,
         output_stream=self.output_streams,
-                  options=task_subgraph_options)
-          ],
+    )
+
+    if hasattr(task_options_proto, 'ext'):
+      # Use the extension mechanism for task_subgraph_options (proto2)
+      task_subgraph_options = calculator_options_pb2.CalculatorOptions()
+      task_subgraph_options.Extensions[task_options_proto.ext].CopyFrom(
+          task_options_proto
+      )
+      node_config.options.CopyFrom(task_subgraph_options)
+    else:
+      # Use the Any type for task_subgraph_options (proto3)
+      task_subgraph_options = any_pb2.Any()
+      task_subgraph_options.Pack(self.task_options.to_pb2())
+      node_config.node_options.append(task_subgraph_options)
+
+    if not enable_flow_limiting:
+      return calculator_pb2.CalculatorGraphConfig(
+          node=[node_config],
           input_stream=self.input_streams,
-          output_stream=self.output_streams)
+          output_stream=self.output_streams,
+      )
     # When a FlowLimiterCalculator is inserted to lower the overall graph
     # latency, the task doesn't guarantee that each input must have the
     # corresponding output.
@@ -120,13 +132,8 @@ class TaskInfo:
         ],
         options=flow_limiter_options)
     config = calculator_pb2.CalculatorGraphConfig(
-        node=[
-            calculator_pb2.CalculatorGraphConfig.Node(
-                calculator=self.task_graph,
-                input_stream=task_subgraph_inputs,
-                output_stream=self.output_streams,
-                options=task_subgraph_options), flow_limiter
-        ],
+        node=[node_config, flow_limiter],
         input_stream=self.input_streams,
-        output_stream=self.output_streams)
+        output_stream=self.output_streams,
+    )
     return config

mediapipe/tasks/python/test/vision/BUILD

@@ -194,6 +194,27 @@ py_test(
     ],
 )
 
+py_test(
+    name = "holistic_landmarker_test",
+    srcs = ["holistic_landmarker_test.py"],
+    data = [
+        "//mediapipe/tasks/testdata/vision:test_images",
+        "//mediapipe/tasks/testdata/vision:test_models",
+        "//mediapipe/tasks/testdata/vision:test_protos",
+    ],
+    tags = ["not_run:arm"],
+    deps = [
+        "//mediapipe/python:_framework_bindings",
+        "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_result_py_pb2",
+        "//mediapipe/tasks/python/core:base_options",
+        "//mediapipe/tasks/python/test:test_utils",
+        "//mediapipe/tasks/python/vision:holistic_landmarker",
+        "//mediapipe/tasks/python/vision/core:image_processing_options",
+        "//mediapipe/tasks/python/vision/core:vision_task_running_mode",
+        "@com_google_protobuf//:protobuf_python",
+    ],
+)
+
 py_test(
     name = "face_aligner_test",
     srcs = ["face_aligner_test.py"],

mediapipe/tasks/python/test/vision/holistic_landmarker_test.py

@@ -0,0 +1,544 @@
+# Copyright 2023 The MediaPipe Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Tests for holistic landmarker."""
+
+import enum
+from unittest import mock
+
+from absl.testing import absltest
+from absl.testing import parameterized
+import numpy as np
+
+from google.protobuf import text_format
+from mediapipe.python._framework_bindings import image as image_module
+from mediapipe.tasks.cc.vision.holistic_landmarker.proto import holistic_result_pb2
+from mediapipe.tasks.python.core import base_options as base_options_module
+from mediapipe.tasks.python.test import test_utils
+from mediapipe.tasks.python.vision import holistic_landmarker
+from mediapipe.tasks.python.vision.core import image_processing_options as image_processing_options_module
+from mediapipe.tasks.python.vision.core import vision_task_running_mode as running_mode_module
+
+
+HolisticLandmarkerResult = holistic_landmarker.HolisticLandmarkerResult
+_HolisticResultProto = holistic_result_pb2.HolisticResult
+_BaseOptions = base_options_module.BaseOptions
+_Image = image_module.Image
+_HolisticLandmarker = holistic_landmarker.HolisticLandmarker
+_HolisticLandmarkerOptions = holistic_landmarker.HolisticLandmarkerOptions
+_RUNNING_MODE = running_mode_module.VisionTaskRunningMode
+_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions
+
+_HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE = 'holistic_landmarker.task'
+_POSE_IMAGE = 'male_full_height_hands.jpg'
+_CAT_IMAGE = 'cat.jpg'
+_EXPECTED_HOLISTIC_RESULT = 'male_full_height_hands_result_cpu.pbtxt'
+_IMAGE_WIDTH = 638
+_IMAGE_HEIGHT = 1000
+_LANDMARKS_MARGIN = 0.03
+_BLENDSHAPES_MARGIN = 0.13
+_VIDEO_LANDMARKS_MARGIN = 0.03
+_VIDEO_BLENDSHAPES_MARGIN = 0.31
+_LIVE_STREAM_LANDMARKS_MARGIN = 0.03
+_LIVE_STREAM_BLENDSHAPES_MARGIN = 0.31
+
+
+def _get_expected_holistic_landmarker_result(
+    file_path: str,
+) -> HolisticLandmarkerResult:
+  holistic_result_file_path = test_utils.get_test_data_path(file_path)
+  with open(holistic_result_file_path, 'rb') as f:
+    holistic_result_proto = _HolisticResultProto()
+    # Use this if a .pb file is available.
+    # holistic_result_proto.ParseFromString(f.read())
+    text_format.Parse(f.read(), holistic_result_proto)
+    holistic_landmarker_result = HolisticLandmarkerResult.create_from_pb2(
+        holistic_result_proto
+    )
+  return holistic_landmarker_result
+
+
+class ModelFileType(enum.Enum):
+  FILE_CONTENT = 1
+  FILE_NAME = 2
+
+
+class HolisticLandmarkerTest(parameterized.TestCase):
+
+  def setUp(self):
+    super().setUp()
+    self.test_image = _Image.create_from_file(
+        test_utils.get_test_data_path(_POSE_IMAGE)
+    )
+    self.model_path = test_utils.get_test_data_path(
+        _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE
+    )
+
+  def _expect_landmarks_correct(
+      self, actual_landmarks, expected_landmarks, margin
+  ):
+    # Expects to have the same number of landmarks detected.
+    self.assertLen(actual_landmarks, len(expected_landmarks))
+
+    for i, elem in enumerate(actual_landmarks):
+      self.assertAlmostEqual(elem.x, expected_landmarks[i].x, delta=margin)
+      self.assertAlmostEqual(elem.y, expected_landmarks[i].y, delta=margin)
+
+  def _expect_blendshapes_correct(
+      self, actual_blendshapes, expected_blendshapes, margin
+  ):
+    # Expects to have the same number of blendshapes.
+    self.assertLen(actual_blendshapes, len(expected_blendshapes))
+
+    for i, elem in enumerate(actual_blendshapes):
+      self.assertEqual(elem.index, expected_blendshapes[i].index)
+      self.assertEqual(
+          elem.category_name, expected_blendshapes[i].category_name
+      )
+      self.assertAlmostEqual(
+          elem.score,
+          expected_blendshapes[i].score,
+          delta=margin,
+      )
+
+  def _expect_holistic_landmarker_results_correct(
+      self,
+      actual_result: HolisticLandmarkerResult,
+      expected_result: HolisticLandmarkerResult,
+      output_segmentation_mask: bool,
+      landmarks_margin: float,
+      blendshapes_margin: float,
+  ):
+    self._expect_landmarks_correct(
+        actual_result.pose_landmarks,
+        expected_result.pose_landmarks,
+        landmarks_margin,
+    )
+    self._expect_landmarks_correct(
+        actual_result.face_landmarks,
+        expected_result.face_landmarks,
+        landmarks_margin,
+    )
+    self._expect_blendshapes_correct(
+        actual_result.face_blendshapes,
+        expected_result.face_blendshapes,
+        blendshapes_margin,
+    )
+    if output_segmentation_mask:
+      self.assertIsInstance(actual_result.segmentation_mask, _Image)
+      self.assertEqual(actual_result.segmentation_mask.width, _IMAGE_WIDTH)
+      self.assertEqual(actual_result.segmentation_mask.height, _IMAGE_HEIGHT)
+    else:
+      self.assertIsNone(actual_result.segmentation_mask)
+
+  def test_create_from_file_succeeds_with_valid_model_path(self):
+    # Creates with default option and valid model file successfully.
+    with _HolisticLandmarker.create_from_model_path(
+        self.model_path
+    ) as landmarker:
+      self.assertIsInstance(landmarker, _HolisticLandmarker)
+
+  def test_create_from_options_succeeds_with_valid_model_path(self):
+    # Creates with options containing model file successfully.
+    base_options = _BaseOptions(model_asset_path=self.model_path)
+    options = _HolisticLandmarkerOptions(base_options=base_options)
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      self.assertIsInstance(landmarker, _HolisticLandmarker)
+
+  def test_create_from_options_fails_with_invalid_model_path(self):
+    # Invalid empty model path.
+    with self.assertRaisesRegex(
+        RuntimeError, 'Unable to open file at /path/to/invalid/model.tflite'
+    ):
+      base_options = _BaseOptions(
+          model_asset_path='/path/to/invalid/model.tflite'
+      )
+      options = _HolisticLandmarkerOptions(base_options=base_options)
+      _HolisticLandmarker.create_from_options(options)
+
+  def test_create_from_options_succeeds_with_valid_model_content(self):
+    # Creates with options containing model content successfully.
+    with open(self.model_path, 'rb') as f:
+      base_options = _BaseOptions(model_asset_buffer=f.read())
+      options = _HolisticLandmarkerOptions(base_options=base_options)
+      landmarker = _HolisticLandmarker.create_from_options(options)
+      self.assertIsInstance(landmarker, _HolisticLandmarker)
+
+  @parameterized.parameters(
+      (
+          ModelFileType.FILE_NAME,
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          False,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+      (
+          ModelFileType.FILE_CONTENT,
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          False,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+      (
+          ModelFileType.FILE_NAME,
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          True,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+      (
+          ModelFileType.FILE_CONTENT,
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          True,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+  )
+  def test_detect(
+      self,
+      model_file_type,
+      model_name,
+      output_segmentation_mask,
+      expected_holistic_landmarker_result,
+  ):
+    # Creates holistic landmarker.
+    model_path = test_utils.get_test_data_path(model_name)
+    if model_file_type is ModelFileType.FILE_NAME:
+      base_options = _BaseOptions(model_asset_path=model_path)
+    elif model_file_type is ModelFileType.FILE_CONTENT:
+      with open(model_path, 'rb') as f:
+        model_content = f.read()
+      base_options = _BaseOptions(model_asset_buffer=model_content)
+    else:
+      # Should never happen
+      raise ValueError('model_file_type is invalid.')
+
+    options = _HolisticLandmarkerOptions(
+        base_options=base_options,
+        output_face_blendshapes=True
+        if expected_holistic_landmarker_result.face_blendshapes
+        else False,
+        output_segmentation_mask=output_segmentation_mask,
+    )
+    landmarker = _HolisticLandmarker.create_from_options(options)
+
+    # Performs holistic landmarks detection on the input.
+    detection_result = landmarker.detect(self.test_image)
+    self._expect_holistic_landmarker_results_correct(
+        detection_result,
+        expected_holistic_landmarker_result,
+        output_segmentation_mask,
+        _LANDMARKS_MARGIN,
+        _BLENDSHAPES_MARGIN,
+    )
+    # Closes the holistic landmarker explicitly when the holistic landmarker is
+    # not used in a context.
+    landmarker.close()
+
+  @parameterized.parameters(
+      (
+          ModelFileType.FILE_NAME,
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          False,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+      (
+          ModelFileType.FILE_CONTENT,
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          True,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+  )
+  def test_detect_in_context(
+      self,
+      model_file_type,
+      model_name,
+      output_segmentation_mask,
+      expected_holistic_landmarker_result,
+  ):
+    # Creates holistic landmarker.
+    model_path = test_utils.get_test_data_path(model_name)
+    if model_file_type is ModelFileType.FILE_NAME:
+      base_options = _BaseOptions(model_asset_path=model_path)
+    elif model_file_type is ModelFileType.FILE_CONTENT:
+      with open(model_path, 'rb') as f:
+        model_content = f.read()
+      base_options = _BaseOptions(model_asset_buffer=model_content)
+    else:
+      # Should never happen
+      raise ValueError('model_file_type is invalid.')
+
+    options = _HolisticLandmarkerOptions(
+        base_options=base_options,
+        output_face_blendshapes=True
+        if expected_holistic_landmarker_result.face_blendshapes
+        else False,
+        output_segmentation_mask=output_segmentation_mask,
+    )
+
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      # Performs holistic landmarks detection on the input.
+      detection_result = landmarker.detect(self.test_image)
+      self._expect_holistic_landmarker_results_correct(
+          detection_result,
+          expected_holistic_landmarker_result,
+          output_segmentation_mask,
+          _LANDMARKS_MARGIN,
+          _BLENDSHAPES_MARGIN,
+      )
+
+  def test_empty_detection_outputs(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path)
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      # Load the cat image.
+      cat_test_image = _Image.create_from_file(
+          test_utils.get_test_data_path(_CAT_IMAGE)
+      )
+      # Performs holistic landmarks detection on the input.
+      detection_result = landmarker.detect(cat_test_image)
+      self.assertEmpty(detection_result.face_landmarks)
+      self.assertEmpty(detection_result.pose_landmarks)
+      self.assertEmpty(detection_result.pose_world_landmarks)
+      self.assertEmpty(detection_result.left_hand_landmarks)
+      self.assertEmpty(detection_result.left_hand_world_landmarks)
+      self.assertEmpty(detection_result.right_hand_landmarks)
+      self.assertEmpty(detection_result.right_hand_world_landmarks)
+      self.assertIsNone(detection_result.face_blendshapes)
+      self.assertIsNone(detection_result.segmentation_mask)
+
+  def test_missing_result_callback(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+    )
+    with self.assertRaisesRegex(
+        ValueError, r'result callback must be provided'
+    ):
+      with _HolisticLandmarker.create_from_options(
+          options
+      ) as unused_landmarker:
+        pass
+
+  @parameterized.parameters((_RUNNING_MODE.IMAGE), (_RUNNING_MODE.VIDEO))
+  def test_illegal_result_callback(self, running_mode):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=running_mode,
+        result_callback=mock.MagicMock(),
+    )
+    with self.assertRaisesRegex(
+        ValueError, r'result callback should not be provided'
+    ):
+      with _HolisticLandmarker.create_from_options(
+          options
+      ) as unused_landmarker:
+        pass
+
+  def test_calling_detect_for_video_in_image_mode(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.IMAGE,
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the video mode'
+      ):
+        landmarker.detect_for_video(self.test_image, 0)
+
+  def test_calling_detect_async_in_image_mode(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.IMAGE,
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the live stream mode'
+      ):
+        landmarker.detect_async(self.test_image, 0)
+
+  def test_calling_detect_in_video_mode(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO,
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the image mode'
+      ):
+        landmarker.detect(self.test_image)
+
+  def test_calling_detect_async_in_video_mode(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO,
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the live stream mode'
+      ):
+        landmarker.detect_async(self.test_image, 0)
+
+  def test_detect_for_video_with_out_of_order_timestamp(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO,
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      unused_result = landmarker.detect_for_video(self.test_image, 1)
+      with self.assertRaisesRegex(
+          ValueError, r'Input timestamp must be monotonically increasing'
+      ):
+        landmarker.detect_for_video(self.test_image, 0)
+
+  @parameterized.parameters(
+      (
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          False,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+      (
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          True,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+  )
+  def test_detect_for_video(
+      self,
+      model_name,
+      output_segmentation_mask,
+      expected_holistic_landmarker_result,
+  ):
+    # Creates holistic landmarker.
+    model_path = test_utils.get_test_data_path(model_name)
+    base_options = _BaseOptions(model_asset_path=model_path)
+    options = _HolisticLandmarkerOptions(
+        base_options=base_options,
+        running_mode=_RUNNING_MODE.VIDEO,
+        output_face_blendshapes=True
+        if expected_holistic_landmarker_result.face_blendshapes
+        else False,
+        output_segmentation_mask=output_segmentation_mask,
+    )
+
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      for timestamp in range(0, 300, 30):
+        # Performs holistic landmarks detection on the input.
+        detection_result = landmarker.detect_for_video(
+            self.test_image, timestamp
+        )
+        # Comparing results.
+        self._expect_holistic_landmarker_results_correct(
+            detection_result,
+            expected_holistic_landmarker_result,
+            output_segmentation_mask,
+            _VIDEO_LANDMARKS_MARGIN,
+            _VIDEO_BLENDSHAPES_MARGIN,
+        )
+
+  def test_calling_detect_in_live_stream_mode(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock(),
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the image mode'
+      ):
+        landmarker.detect(self.test_image)
+
+  def test_calling_detect_for_video_in_live_stream_mode(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock(),
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the video mode'
+      ):
+        landmarker.detect_for_video(self.test_image, 0)
+
+  def test_detect_async_calls_with_illegal_timestamp(self):
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock(),
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      landmarker.detect_async(self.test_image, 100)
+      with self.assertRaisesRegex(
+          ValueError, r'Input timestamp must be monotonically increasing'
+      ):
+        landmarker.detect_async(self.test_image, 0)
+
+  @parameterized.parameters(
+      (
+          _POSE_IMAGE,
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          False,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+      (
+          _POSE_IMAGE,
+          _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE,
+          True,
+          _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT),
+      ),
+  )
+  def test_detect_async_calls(
+      self,
+      image_path,
+      model_name,
+      output_segmentation_mask,
+      expected_holistic_landmarker_result,
+  ):
+    test_image = _Image.create_from_file(
+        test_utils.get_test_data_path(image_path)
+    )
+    observed_timestamp_ms = -1
+
+    def check_result(
+        result: HolisticLandmarkerResult,
+        output_image: _Image,
+        timestamp_ms: int,
+    ):
+      # Comparing results.
+      self._expect_holistic_landmarker_results_correct(
+          result,
+          expected_holistic_landmarker_result,
+          output_segmentation_mask,
+          _LIVE_STREAM_LANDMARKS_MARGIN,
+          _LIVE_STREAM_BLENDSHAPES_MARGIN,
+      )
+      self.assertTrue(
+          np.array_equal(output_image.numpy_view(), test_image.numpy_view())
+      )
+      self.assertLess(observed_timestamp_ms, timestamp_ms)
+      self.observed_timestamp_ms = timestamp_ms
+
+    model_path = test_utils.get_test_data_path(model_name)
+    options = _HolisticLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        output_face_blendshapes=True
+        if expected_holistic_landmarker_result.face_blendshapes
+        else False,
+        output_segmentation_mask=output_segmentation_mask,
+        result_callback=check_result,
+    )
+    with _HolisticLandmarker.create_from_options(options) as landmarker:
+      for timestamp in range(0, 300, 30):
+        landmarker.detect_async(test_image, timestamp)
+
+
+if __name__ == '__main__':
+  absltest.main()

mediapipe/tasks/python/vision/BUILD

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-# Placeholder: load py_library
 # Placeholder for internal Python strict library and test compatibility macro.
 
 package(default_visibility = ["//visibility:public"])
@@ -243,6 +242,30 @@ py_library(
     ],
 )
 
+py_library(
+    name = "holistic_landmarker",
+    srcs = [
+        "holistic_landmarker.py",
+    ],
+    deps = [
+        "//mediapipe/framework/formats:classification_py_pb2",
+        "//mediapipe/framework/formats:landmark_py_pb2",
+        "//mediapipe/python:_framework_bindings",
+        "//mediapipe/python:packet_creator",
+        "//mediapipe/python:packet_getter",
+        "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_landmarker_graph_options_py_pb2",
+        "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_result_py_pb2",
+        "//mediapipe/tasks/python/components/containers:category",
+        "//mediapipe/tasks/python/components/containers:landmark",
+        "//mediapipe/tasks/python/core:base_options",
+        "//mediapipe/tasks/python/core:optional_dependencies",
+        "//mediapipe/tasks/python/core:task_info",
+        "//mediapipe/tasks/python/vision/core:base_vision_task_api",
+        "//mediapipe/tasks/python/vision/core:image_processing_options",
+        "//mediapipe/tasks/python/vision/core:vision_task_running_mode",
+    ],
+)
+
 py_library(
     name = "face_stylizer",
     srcs = [

mediapipe/tasks/python/vision/holistic_landmarker.py

@@ -0,0 +1,576 @@
+# Copyright 2023 The MediaPipe Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""MediaPipe holistic landmarker task."""
+
+import dataclasses
+from typing import Callable, List, Mapping, Optional
+
+from mediapipe.framework.formats import classification_pb2
+from mediapipe.framework.formats import landmark_pb2
+from mediapipe.python import packet_creator
+from mediapipe.python import packet_getter
+from mediapipe.python._framework_bindings import image as image_module
+from mediapipe.python._framework_bindings import packet as packet_module
+from mediapipe.tasks.cc.vision.holistic_landmarker.proto import holistic_landmarker_graph_options_pb2
+from mediapipe.tasks.cc.vision.holistic_landmarker.proto import holistic_result_pb2
+from mediapipe.tasks.python.components.containers import category as category_module
+from mediapipe.tasks.python.components.containers import landmark as landmark_module
+from mediapipe.tasks.python.core import base_options as base_options_module
+from mediapipe.tasks.python.core import task_info as task_info_module
+from mediapipe.tasks.python.core.optional_dependencies import doc_controls
+from mediapipe.tasks.python.vision.core import base_vision_task_api
+from mediapipe.tasks.python.vision.core import image_processing_options as image_processing_options_module
+from mediapipe.tasks.python.vision.core import vision_task_running_mode as running_mode_module
+
+_BaseOptions = base_options_module.BaseOptions
+_HolisticResultProto = holistic_result_pb2.HolisticResult
+_HolisticLandmarkerGraphOptionsProto = (
+    holistic_landmarker_graph_options_pb2.HolisticLandmarkerGraphOptions
+)
+_RunningMode = running_mode_module.VisionTaskRunningMode
+_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions
+_TaskInfo = task_info_module.TaskInfo
+
+_IMAGE_IN_STREAM_NAME = 'image_in'
+_IMAGE_OUT_STREAM_NAME = 'image_out'
+_IMAGE_TAG = 'IMAGE'
+
+_POSE_LANDMARKS_STREAM_NAME = 'pose_landmarks'
+_POSE_LANDMARKS_TAG_NAME = 'POSE_LANDMARKS'
+_POSE_WORLD_LANDMARKS_STREAM_NAME = 'pose_world_landmarks'
+_POSE_WORLD_LANDMARKS_TAG = 'POSE_WORLD_LANDMARKS'
+_POSE_SEGMENTATION_MASK_STREAM_NAME = 'pose_segmentation_mask'
+_POSE_SEGMENTATION_MASK_TAG = 'POSE_SEGMENTATION_MASK'
+_FACE_LANDMARKS_STREAM_NAME = 'face_landmarks'
+_FACE_LANDMARKS_TAG = 'FACE_LANDMARKS'
+_FACE_BLENDSHAPES_STREAM_NAME = 'extra_blendshapes'
+_FACE_BLENDSHAPES_TAG = 'FACE_BLENDSHAPES'
+_LEFT_HAND_LANDMARKS_STREAM_NAME = 'left_hand_landmarks'
+_LEFT_HAND_LANDMARKS_TAG = 'LEFT_HAND_LANDMARKS'
+_LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME = 'left_hand_world_landmarks'
+_LEFT_HAND_WORLD_LANDMARKS_TAG = 'LEFT_HAND_WORLD_LANDMARKS'
+_RIGHT_HAND_LANDMARKS_STREAM_NAME = 'right_hand_landmarks'
+_RIGHT_HAND_LANDMARKS_TAG = 'RIGHT_HAND_LANDMARKS'
+_RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME = 'right_hand_world_landmarks'
+_RIGHT_HAND_WORLD_LANDMARKS_TAG = 'RIGHT_HAND_WORLD_LANDMARKS'
+
+_TASK_GRAPH_NAME = (
+    'mediapipe.tasks.vision.holistic_landmarker.HolisticLandmarkerGraph'
+)
+_MICRO_SECONDS_PER_MILLISECOND = 1000
+
+
+@dataclasses.dataclass
+class HolisticLandmarkerResult:
+  """The holistic landmarks result from HolisticLandmarker, where each vector element represents a single holistic detected in the image.
+
+  Attributes:
+    face_landmarks: Detected face landmarks in normalized image coordinates.
+    pose_landmarks: Detected pose landmarks in normalized image coordinates.
+    pose_world_landmarks: Detected pose world landmarks in image coordinates.
+    left_hand_landmarks: Detected left hand landmarks in normalized image
+      coordinates.
+    left_hand_world_landmarks: Detected left hand landmarks in image
+      coordinates.
+    right_hand_landmarks: Detected right hand landmarks in normalized image
+      coordinates.
+    right_hand_world_landmarks: Detected right hand landmarks in image
+      coordinates.
+    face_blendshapes: Optional face blendshapes.
+    segmentation_mask: Optional segmentation mask for pose.
+  """
+
+  face_landmarks: List[landmark_module.NormalizedLandmark]
+  pose_landmarks: List[landmark_module.NormalizedLandmark]
+  pose_world_landmarks: List[landmark_module.Landmark]
+  left_hand_landmarks: List[landmark_module.NormalizedLandmark]
+  left_hand_world_landmarks: List[landmark_module.Landmark]
+  right_hand_landmarks: List[landmark_module.NormalizedLandmark]
+  right_hand_world_landmarks: List[landmark_module.Landmark]
+  face_blendshapes: Optional[List[category_module.Category]] = None
+  segmentation_mask: Optional[image_module.Image] = None
+
+  @classmethod
+  @doc_controls.do_not_generate_docs
+  def create_from_pb2(
+      cls, pb2_obj: _HolisticResultProto
+  ) -> 'HolisticLandmarkerResult':
+    """Creates a `HolisticLandmarkerResult` object from the given protobuf object."""
+    face_blendshapes = None
+    if hasattr(pb2_obj, 'face_blendshapes'):
+      face_blendshapes = [
+          category_module.Category(
+              score=classification.score,
+              index=classification.index,
+              category_name=classification.label,
+              display_name=classification.display_name,
+          )
+          for classification in pb2_obj.face_blendshapes.classification
+      ]
+
+    return HolisticLandmarkerResult(
+        face_landmarks=[
+            landmark_module.NormalizedLandmark.create_from_pb2(landmark)
+            for landmark in pb2_obj.face_landmarks.landmark
+        ],
+        pose_landmarks=[
+            landmark_module.NormalizedLandmark.create_from_pb2(landmark)
+            for landmark in pb2_obj.pose_landmarks.landmark
+        ],
+        pose_world_landmarks=[
+            landmark_module.Landmark.create_from_pb2(landmark)
+            for landmark in pb2_obj.pose_world_landmarks.landmark
+        ],
+        left_hand_landmarks=[
+            landmark_module.NormalizedLandmark.create_from_pb2(landmark)
+            for landmark in pb2_obj.left_hand_landmarks.landmark
+        ],
+        left_hand_world_landmarks=[],
+        right_hand_landmarks=[
+            landmark_module.NormalizedLandmark.create_from_pb2(landmark)
+            for landmark in pb2_obj.right_hand_landmarks.landmark
+        ],
+        right_hand_world_landmarks=[],
+        face_blendshapes=face_blendshapes,
+        segmentation_mask=None,
+    )
+
+
+def _build_landmarker_result(
+    output_packets: Mapping[str, packet_module.Packet]
+) -> HolisticLandmarkerResult:
+  """Constructs a `HolisticLandmarksDetectionResult` from output packets."""
+  holistic_landmarker_result = HolisticLandmarkerResult(
+      [], [], [], [], [], [], []
+  )
+
+  face_landmarks_proto_list = packet_getter.get_proto(
+      output_packets[_FACE_LANDMARKS_STREAM_NAME]
+  )
+
+  pose_landmarks_proto_list = packet_getter.get_proto(
+      output_packets[_POSE_LANDMARKS_STREAM_NAME]
+  )
+
+  pose_world_landmarks_proto_list = packet_getter.get_proto(
+      output_packets[_POSE_WORLD_LANDMARKS_STREAM_NAME]
+  )
+
+  left_hand_landmarks_proto_list = packet_getter.get_proto(
+      output_packets[_LEFT_HAND_LANDMARKS_STREAM_NAME]
+  )
+
+  left_hand_world_landmarks_proto_list = packet_getter.get_proto(
+      output_packets[_LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME]
+  )
+
+  right_hand_landmarks_proto_list = packet_getter.get_proto(
+      output_packets[_RIGHT_HAND_LANDMARKS_STREAM_NAME]
+  )
+
+  right_hand_world_landmarks_proto_list = packet_getter.get_proto(
+      output_packets[_RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME]
+  )
+
+  face_landmarks = landmark_pb2.NormalizedLandmarkList()
+  face_landmarks.MergeFrom(face_landmarks_proto_list)
+  for face_landmark in face_landmarks.landmark:
+    holistic_landmarker_result.face_landmarks.append(
+        landmark_module.NormalizedLandmark.create_from_pb2(face_landmark)
+    )
+
+  pose_landmarks = landmark_pb2.NormalizedLandmarkList()
+  pose_landmarks.MergeFrom(pose_landmarks_proto_list)
+  for pose_landmark in pose_landmarks.landmark:
+    holistic_landmarker_result.pose_landmarks.append(
+        landmark_module.NormalizedLandmark.create_from_pb2(pose_landmark)
+    )
+
+  pose_world_landmarks = landmark_pb2.LandmarkList()
+  pose_world_landmarks.MergeFrom(pose_world_landmarks_proto_list)
+  for pose_world_landmark in pose_world_landmarks.landmark:
+    holistic_landmarker_result.pose_world_landmarks.append(
+        landmark_module.Landmark.create_from_pb2(pose_world_landmark)
+    )
+
+  left_hand_landmarks = landmark_pb2.NormalizedLandmarkList()
+  left_hand_landmarks.MergeFrom(left_hand_landmarks_proto_list)
+  for hand_landmark in left_hand_landmarks.landmark:
+    holistic_landmarker_result.left_hand_landmarks.append(
+        landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark)
+    )
+
+  left_hand_world_landmarks = landmark_pb2.LandmarkList()
+  left_hand_world_landmarks.MergeFrom(left_hand_world_landmarks_proto_list)
+  for left_hand_world_landmark in left_hand_world_landmarks.landmark:
+    holistic_landmarker_result.left_hand_world_landmarks.append(
+        landmark_module.Landmark.create_from_pb2(left_hand_world_landmark)
+    )
+
+  right_hand_landmarks = landmark_pb2.NormalizedLandmarkList()
+  right_hand_landmarks.MergeFrom(right_hand_landmarks_proto_list)
+  for hand_landmark in right_hand_landmarks.landmark:
+    holistic_landmarker_result.right_hand_landmarks.append(
+        landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark)
+    )
+
+  right_hand_world_landmarks = landmark_pb2.LandmarkList()
+  right_hand_world_landmarks.MergeFrom(right_hand_world_landmarks_proto_list)
+  for right_hand_world_landmark in right_hand_world_landmarks.landmark:
+    holistic_landmarker_result.right_hand_world_landmarks.append(
+        landmark_module.Landmark.create_from_pb2(right_hand_world_landmark)
+    )
+
+  if _FACE_BLENDSHAPES_STREAM_NAME in output_packets:
+    face_blendshapes_proto_list = packet_getter.get_proto(
+        output_packets[_FACE_BLENDSHAPES_STREAM_NAME]
+    )
+    face_blendshapes_classifications = classification_pb2.ClassificationList()
+    face_blendshapes_classifications.MergeFrom(face_blendshapes_proto_list)
+    holistic_landmarker_result.face_blendshapes = []
+    for face_blendshapes in face_blendshapes_classifications.classification:
+      holistic_landmarker_result.face_blendshapes.append(
+          category_module.Category(
+              index=face_blendshapes.index,
+              score=face_blendshapes.score,
+              display_name=face_blendshapes.display_name,
+              category_name=face_blendshapes.label,
+          )
+      )
+
+  if _POSE_SEGMENTATION_MASK_STREAM_NAME in output_packets:
+    holistic_landmarker_result.segmentation_mask = packet_getter.get_image(
+        output_packets[_POSE_SEGMENTATION_MASK_STREAM_NAME]
+    )
+
+  return holistic_landmarker_result
+
+
+@dataclasses.dataclass
+class HolisticLandmarkerOptions:
+  """Options for the holistic landmarker task.
+
+  Attributes:
+    base_options: Base options for the holistic landmarker task.
+    running_mode: The running mode of the task. Default to the image mode.
+      HolisticLandmarker has three running modes: 1) The image mode for
+      detecting holistic landmarks on single image inputs. 2) The video mode for
+      detecting holistic landmarks on the decoded frames of a video. 3) The live
+      stream mode for detecting holistic landmarks on the live stream of input
+      data, such as from camera. In this mode, the "result_callback" below must
+      be specified to receive the detection results asynchronously.
+    min_face_detection_confidence: The minimum confidence score for the face
+      detection to be considered successful.
+    min_face_suppression_threshold: The minimum non-maximum-suppression
+      threshold for face detection to be considered overlapped.
+    min_face_landmarks_confidence: The minimum confidence score for the face
+      landmark detection to be considered successful.
+    min_pose_detection_confidence: The minimum confidence score for the pose
+      detection to be considered successful.
+    min_pose_suppression_threshold: The minimum non-maximum-suppression
+      threshold for pose detection to be considered overlapped.
+    min_pose_landmarks_confidence: The minimum confidence score for the pose
+      landmark detection to be considered successful.
+    min_hand_landmarks_confidence: The minimum confidence score for the hand
+      landmark detection to be considered successful.
+    output_face_blendshapes: Whether HolisticLandmarker outputs face blendshapes
+      classification. Face blendshapes are used for rendering the 3D face model.
+    output_segmentation_mask: whether to output segmentation masks.
+    result_callback: The user-defined result callback for processing live stream
+      data. The result callback should only be specified when the running mode
+      is set to the live stream mode.
+  """
+
+  base_options: _BaseOptions
+  running_mode: _RunningMode = _RunningMode.IMAGE
+  min_face_detection_confidence: float = 0.5
+  min_face_suppression_threshold: float = 0.5
+  min_face_landmarks_confidence: float = 0.5
+  min_pose_detection_confidence: float = 0.5
+  min_pose_suppression_threshold: float = 0.5
+  min_pose_landmarks_confidence: float = 0.5
+  min_hand_landmarks_confidence: float = 0.5
+  output_face_blendshapes: bool = False
+  output_segmentation_mask: bool = False
+  result_callback: Optional[
+      Callable[[HolisticLandmarkerResult, image_module.Image, int], None]
+  ] = None
+
+  @doc_controls.do_not_generate_docs
+  def to_pb2(self) -> _HolisticLandmarkerGraphOptionsProto:
+    """Generates an HolisticLandmarkerGraphOptions protobuf object."""
+    base_options_proto = self.base_options.to_pb2()
+    base_options_proto.use_stream_mode = (
+        False if self.running_mode == _RunningMode.IMAGE else True
+    )
+
+    # Initialize the holistic landmarker options from base options.
+    holistic_landmarker_options_proto = _HolisticLandmarkerGraphOptionsProto(
+        base_options=base_options_proto
+    )
+    # Configure face detector and face landmarks detector options.
+    holistic_landmarker_options_proto.face_detector_graph_options.min_detection_confidence = (
+        self.min_face_detection_confidence
+    )
+    holistic_landmarker_options_proto.face_detector_graph_options.min_suppression_threshold = (
+        self.min_face_suppression_threshold
+    )
+    holistic_landmarker_options_proto.face_landmarks_detector_graph_options.min_detection_confidence = (
+        self.min_face_landmarks_confidence
+    )
+    # Configure pose detector and pose landmarks detector options.
+    holistic_landmarker_options_proto.pose_detector_graph_options.min_detection_confidence = (
+        self.min_pose_detection_confidence
+    )
+    holistic_landmarker_options_proto.pose_detector_graph_options.min_suppression_threshold = (
+        self.min_pose_suppression_threshold
+    )
+    holistic_landmarker_options_proto.pose_landmarks_detector_graph_options.min_detection_confidence = (
+        self.min_pose_landmarks_confidence
+    )
+    # Configure hand landmarks detector options.
+    holistic_landmarker_options_proto.hand_landmarks_detector_graph_options.min_detection_confidence = (
+        self.min_hand_landmarks_confidence
+    )
+    return holistic_landmarker_options_proto
+
+
+class HolisticLandmarker(base_vision_task_api.BaseVisionTaskApi):
+  """Class that performs holistic landmarks detection on images."""
+
+  @classmethod
+  def create_from_model_path(cls, model_path: str) -> 'HolisticLandmarker':
+    """Creates an `HolisticLandmarker` object from a TensorFlow Lite model and the default `HolisticLandmarkerOptions`.
+
+    Note that the created `HolisticLandmarker` instance is in image mode, for
+    detecting holistic landmarks on single image inputs.
+
+    Args:
+      model_path: Path to the model.
+
+    Returns:
+      `HolisticLandmarker` object that's created from the model file and the
+      default `HolisticLandmarkerOptions`.
+
+    Raises:
+      ValueError: If failed to create `HolisticLandmarker` object from the
+        provided file such as invalid file path.
+      RuntimeError: If other types of error occurred.
+    """
+    base_options = _BaseOptions(model_asset_path=model_path)
+    options = HolisticLandmarkerOptions(
+        base_options=base_options, running_mode=_RunningMode.IMAGE
+    )
+    return cls.create_from_options(options)
+
+  @classmethod
+  def create_from_options(
+      cls, options: HolisticLandmarkerOptions
+  ) -> 'HolisticLandmarker':
+    """Creates the `HolisticLandmarker` object from holistic landmarker options.
+
+    Args:
+      options: Options for the holistic landmarker task.
+
+    Returns:
+      `HolisticLandmarker` object that's created from `options`.
+
+    Raises:
+      ValueError: If failed to create `HolisticLandmarker` object from
+        `HolisticLandmarkerOptions` such as missing the model.
+      RuntimeError: If other types of error occurred.
+    """
+
+    def packets_callback(output_packets: Mapping[str, packet_module.Packet]):
+      if output_packets[_IMAGE_OUT_STREAM_NAME].is_empty():
+        return
+
+      image = packet_getter.get_image(output_packets[_IMAGE_OUT_STREAM_NAME])
+
+      if output_packets[_FACE_LANDMARKS_STREAM_NAME].is_empty():
+        empty_packet = output_packets[_FACE_LANDMARKS_STREAM_NAME]
+        options.result_callback(
+            HolisticLandmarkerResult([], [], [], [], [], [], []),
+            image,
+            empty_packet.timestamp.value // _MICRO_SECONDS_PER_MILLISECOND,
+        )
+        return
+
+      holistic_landmarks_detection_result = _build_landmarker_result(
+          output_packets
+      )
+      timestamp = output_packets[_FACE_LANDMARKS_STREAM_NAME].timestamp
+      options.result_callback(
+          holistic_landmarks_detection_result,
+          image,
+          timestamp.value // _MICRO_SECONDS_PER_MILLISECOND,
+      )
+
+    output_streams = [
+        ':'.join([_FACE_LANDMARKS_TAG, _FACE_LANDMARKS_STREAM_NAME]),
+        ':'.join([_POSE_LANDMARKS_TAG_NAME, _POSE_LANDMARKS_STREAM_NAME]),
+        ':'.join(
+            [_POSE_WORLD_LANDMARKS_TAG, _POSE_WORLD_LANDMARKS_STREAM_NAME]
+        ),
+        ':'.join([_LEFT_HAND_LANDMARKS_TAG, _LEFT_HAND_LANDMARKS_STREAM_NAME]),
+        ':'.join([
+            _LEFT_HAND_WORLD_LANDMARKS_TAG,
+            _LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME,
+        ]),
+        ':'.join(
+            [_RIGHT_HAND_LANDMARKS_TAG, _RIGHT_HAND_LANDMARKS_STREAM_NAME]
+        ),
+        ':'.join([
+            _RIGHT_HAND_WORLD_LANDMARKS_TAG,
+            _RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME,
+        ]),
+        ':'.join([_IMAGE_TAG, _IMAGE_OUT_STREAM_NAME]),
+    ]
+
+    if options.output_segmentation_mask:
+      output_streams.append(
+          ':'.join(
+              [_POSE_SEGMENTATION_MASK_TAG, _POSE_SEGMENTATION_MASK_STREAM_NAME]
+          )
+      )
+
+    if options.output_face_blendshapes:
+      output_streams.append(
+          ':'.join([_FACE_BLENDSHAPES_TAG, _FACE_BLENDSHAPES_STREAM_NAME])
+      )
+
+    task_info = _TaskInfo(
+        task_graph=_TASK_GRAPH_NAME,
+        input_streams=[
+            ':'.join([_IMAGE_TAG, _IMAGE_IN_STREAM_NAME]),
+        ],
+        output_streams=output_streams,
+        task_options=options,
+    )
+    return cls(
+        task_info.generate_graph_config(
+            enable_flow_limiting=options.running_mode
+            == _RunningMode.LIVE_STREAM
+        ),
+        options.running_mode,
+        packets_callback if options.result_callback else None,
+    )
+
+  def detect(
+      self,
+      image: image_module.Image,
+  ) -> HolisticLandmarkerResult:
+    """Performs holistic landmarks detection on the given image.
+
+    Only use this method when the HolisticLandmarker is created with the image
+    running mode.
+
+    The image can be of any size with format RGB or RGBA.
+
+    Args:
+      image: MediaPipe Image.
+
+    Returns:
+      The holistic landmarks detection results.
+
+    Raises:
+      ValueError: If any of the input arguments is invalid.
+      RuntimeError: If holistic landmarker detection failed to run.
+    """
+    output_packets = self._process_image_data({
+        _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image),
+    })
+
+    if output_packets[_FACE_LANDMARKS_STREAM_NAME].is_empty():
+      return HolisticLandmarkerResult([], [], [], [], [], [], [])
+
+    return _build_landmarker_result(output_packets)
+
+  def detect_for_video(
+      self,
+      image: image_module.Image,
+      timestamp_ms: int,
+  ) -> HolisticLandmarkerResult:
+    """Performs holistic landmarks detection on the provided video frame.
+
+    Only use this method when the HolisticLandmarker is created with the video
+    running mode.
+
+    Only use this method when the HolisticLandmarker is created with the video
+    running mode. It's required to provide the video frame's timestamp (in
+    milliseconds) along with the video frame. The input timestamps should be
+    monotonically increasing for adjacent calls of this method.
+
+    Args:
+      image: MediaPipe Image.
+      timestamp_ms: The timestamp of the input video frame in milliseconds.
+
+    Returns:
+      The holistic landmarks detection results.
+
+    Raises:
+      ValueError: If any of the input arguments is invalid.
+      RuntimeError: If holistic landmarker detection failed to run.
+    """
+    output_packets = self._process_video_data({
+        _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image).at(
+            timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND
+        ),
+    })
+
+    if output_packets[_FACE_LANDMARKS_STREAM_NAME].is_empty():
+      return HolisticLandmarkerResult([], [], [], [], [], [], [])
+
+    return _build_landmarker_result(output_packets)
+
+  def detect_async(
+      self,
+      image: image_module.Image,
+      timestamp_ms: int,
+  ) -> None:
+    """Sends live image data to perform holistic landmarks detection.
+
+    The results will be available via the "result_callback" provided in the
+    HolisticLandmarkerOptions. Only use this method when the HolisticLandmarker
+    is
+    created with the live stream running mode.
+
+    Only use this method when the HolisticLandmarker is created with the live
+    stream running mode. The input timestamps should be monotonically increasing
+    for adjacent calls of this method. This method will return immediately after
+    the input image is accepted. The results will be available via the
+    `result_callback` provided in the `HolisticLandmarkerOptions`. The
+    `detect_async` method is designed to process live stream data such as
+    camera input. To lower the overall latency, holistic landmarker may drop the
+    input images if needed. In other words, it's not guaranteed to have output
+    per input image.
+
+    The `result_callback` provides:
+      - The holistic landmarks detection results.
+      - The input image that the holistic landmarker runs on.
+      - The input timestamp in milliseconds.
+
+    Args:
+      image: MediaPipe Image.
+      timestamp_ms: The timestamp of the input image in milliseconds.
+
+    Raises:
+      ValueError: If the current input timestamp is smaller than what the
+      holistic landmarker has already processed.
+    """
+    self._send_live_stream_data({
+        _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image).at(
+            timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND
+        ),
+    })

mediapipe/tasks/testdata/vision/BUILD

@@ -48,6 +48,7 @@ mediapipe_files(srcs = [
     "face_landmark.tflite",
     "face_landmarker.task",
     "face_landmarker_v2.task",
+    "face_landmarker_v2_with_blendshapes.task",
     "face_stylizer_color_ink.task",
     "fist.jpg",
     "fist.png",
@@ -57,9 +58,11 @@ mediapipe_files(srcs = [
     "hand_landmark_lite.tflite",
     "hand_landmarker.task",
     "handrecrop_2020_07_21_v0.f16.tflite",
+    "holistic_landmarker.task",
     "left_hands.jpg",
     "left_hands_rotated.jpg",
     "leopard_bg_removal_result_512x512.png",
+    "male_full_height_hands.jpg",
     "mobilenet_v1_0.25_192_quantized_1_default_1.tflite",
     "mobilenet_v1_0.25_224_1_default_1.tflite",
     "mobilenet_v1_0.25_224_1_metadata_1.tflite",
@@ -141,6 +144,7 @@ filegroup(
         "left_hands.jpg",
         "left_hands_rotated.jpg",
         "leopard_bg_removal_result_512x512.png",
+        "male_full_height_hands.jpg",
         "mozart_square.jpg",
         "multi_objects.jpg",
         "multi_objects_rotated.jpg",
@@ -185,6 +189,7 @@ filegroup(
         "face_detection_short_range.tflite",
         "face_landmarker.task",
         "face_landmarker_v2.task",
+        "face_landmarker_v2_with_blendshapes.task",
         "face_stylizer_color_ink.task",
         "gesture_recognizer.task",
         "hair_segmentation.tflite",
@@ -192,6 +197,7 @@ filegroup(
         "hand_landmark_lite.tflite",
         "hand_landmarker.task",
         "handrecrop_2020_07_21_v0.f16.tflite",
+        "holistic_landmarker.task",
         "mobilenet_v1_0.25_192_quantized_1_default_1.tflite",
         "mobilenet_v1_0.25_224_1_default_1.tflite",
         "mobilenet_v1_0.25_224_1_metadata_1.tflite",

mediapipe/tasks/web/core/task_runner_test_utils.ts

@@ -71,6 +71,15 @@ export interface MediapipeTasksFake {
 /** An map of field paths to values */
 export type FieldPathToValue = [string[] | string, unknown];
 
+type JsonObject = Record<string, unknown>;
+
+/**
+ * The function to convert a binary proto to a JsonObject.
+ * For example, the deserializer of HolisticLandmarkerOptions's binary proto is
+ * HolisticLandmarkerOptions.deserializeBinary(binaryProto).toObject().
+ */
+export type Deserializer = (binaryProto: Uint8Array) => JsonObject;
+
 /**
  * Verifies that the graph has been initialized and that it contains the
  * provided options.
@@ -79,6 +88,7 @@ export function verifyGraph(
     tasksFake: MediapipeTasksFake,
     expectedCalculatorOptions?: FieldPathToValue,
     expectedBaseOptions?: FieldPathToValue,
+    deserializer?: Deserializer,
     ): void {
   expect(tasksFake.graph).toBeDefined();
   // Our graphs should have at least one node in them for processing, and
@@ -89,22 +99,31 @@ export function verifyGraph(
   expect(node).toEqual(
       jasmine.objectContaining({calculator: tasksFake.calculatorName}));
 
+  let proto;
+  if (deserializer) {
+    const binaryProto =
+    tasksFake.graph!.getNodeList()[0].getNodeOptionsList()[0].getValue() as
+    Uint8Array;
+    proto = deserializer(binaryProto);
+  } else {
+    proto = (node.options as {ext: unknown}).ext;
+  }
+
   if (expectedBaseOptions) {
     const [fieldPath, value] = expectedBaseOptions;
-    let proto = (node.options as {ext: {baseOptions: unknown}}).ext.baseOptions;
+    let baseOptions = (proto as {baseOptions: unknown}).baseOptions;
     for (const fieldName of (
              Array.isArray(fieldPath) ? fieldPath : [fieldPath])) {
-      proto = ((proto ?? {}) as Record<string, unknown>)[fieldName];
+      baseOptions = ((baseOptions ?? {}) as JsonObject)[fieldName];
     }
-    expect(proto).toEqual(value);
+    expect(baseOptions).toEqual(value);
   }
 
   if (expectedCalculatorOptions) {
     const [fieldPath, value] = expectedCalculatorOptions;
-    let proto = (node.options as {ext: unknown}).ext;
     for (const fieldName of (
              Array.isArray(fieldPath) ? fieldPath : [fieldPath])) {
-      proto = ((proto ?? {}) as Record<string, unknown>)[fieldName];
+      proto = ((proto ?? {}) as JsonObject)[fieldName];
     }
     expect(proto).toEqual(value);
   }

mediapipe/tasks/web/vision/BUILD

@@ -27,6 +27,7 @@ VISION_LIBS = [
     "//mediapipe/tasks/web/vision/face_stylizer",
     "//mediapipe/tasks/web/vision/gesture_recognizer",
     "//mediapipe/tasks/web/vision/hand_landmarker",
+    "//mediapipe/tasks/web/vision/holistic_landmarker",
     "//mediapipe/tasks/web/vision/image_classifier",
     "//mediapipe/tasks/web/vision/image_embedder",
     "//mediapipe/tasks/web/vision/image_segmenter",

mediapipe/tasks/web/vision/core/drawing_utils_category_mask.ts

@@ -14,7 +14,7 @@
  * limitations under the License.
  */
 
-import {assertNotNull, MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
+import {assertExists, MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
 import {ImageSource} from '../../../../web/graph_runner/graph_runner';
 
 /**
@@ -159,13 +159,13 @@ export class CategoryMaskShaderContext extends MPImageShaderContext {
   protected override setupShaders(): void {
     super.setupShaders();
     const gl = this.gl!;
-    this.backgroundTextureUniform = assertNotNull(
+    this.backgroundTextureUniform = assertExists(
         gl.getUniformLocation(this.program!, 'backgroundTexture'),
         'Uniform location');
-    this.colorMappingTextureUniform = assertNotNull(
+    this.colorMappingTextureUniform = assertExists(
         gl.getUniformLocation(this.program!, 'colorMappingTexture'),
         'Uniform location');
-    this.maskTextureUniform = assertNotNull(
+    this.maskTextureUniform = assertExists(
         gl.getUniformLocation(this.program!, 'maskTexture'),
         'Uniform location');
   }

mediapipe/tasks/web/vision/core/drawing_utils_confidence_mask.ts

@@ -14,7 +14,7 @@
  * limitations under the License.
  */
 
-import {assertNotNull, MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
+import {assertExists, MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
 import {ImageSource} from '../../../../web/graph_runner/graph_runner';
 
 /**
@@ -60,13 +60,13 @@ export class ConfidenceMaskShaderContext extends MPImageShaderContext {
   protected override setupShaders(): void {
     super.setupShaders();
     const gl = this.gl!;
-    this.defaultTextureUniform = assertNotNull(
+    this.defaultTextureUniform = assertExists(
         gl.getUniformLocation(this.program!, 'defaultTexture'),
         'Uniform location');
-    this.overlayTextureUniform = assertNotNull(
+    this.overlayTextureUniform = assertExists(
         gl.getUniformLocation(this.program!, 'overlayTexture'),
         'Uniform location');
-    this.maskTextureUniform = assertNotNull(
+    this.maskTextureUniform = assertExists(
         gl.getUniformLocation(this.program!, 'maskTexture'),
         'Uniform location');
   }

mediapipe/tasks/web/vision/core/image.ts

@@ -14,7 +14,7 @@
  * limitations under the License.
  */
 
-import {assertNotNull, MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
+import {assertExists, MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
 
 /** Number of instances a user can keep alive before we raise a warning. */
 const INSTANCE_COUNT_WARNING_THRESHOLD = 250;
@@ -249,8 +249,8 @@ export class MPImage {
           'is passed when iniitializing the image.');
     }
     if (!this.gl) {
-      this.gl = assertNotNull(
-          this.canvas.getContext('webgl2'),
+      this.gl = assertExists(
+          this.canvas.getContext('webgl2') as WebGL2RenderingContext,
           'You cannot use a canvas that is already bound to a different ' +
               'type of rendering context.');
     }

mediapipe/tasks/web/vision/core/image_shader_context.ts

@@ -32,9 +32,9 @@ const FRAGMENT_SHADER = `
   }
  `;
 
-/** Helper to assert that `value` is not null.  */
-export function assertNotNull<T>(value: T|null, msg: string): T {
-  if (value === null) {
+/** Helper to assert that `value` is not null or undefined.  */
+export function assertExists<T>(value: T, msg: string): NonNullable<T> {
+  if (!value) {
     throw new Error(`Unable to obtain required WebGL resource: ${msg}`);
   }
   return value;
@@ -105,7 +105,7 @@ export class MPImageShaderContext {
   private compileShader(source: string, type: number): WebGLShader {
     const gl = this.gl!;
     const shader =
-        assertNotNull(gl.createShader(type), 'Failed to create WebGL shader');
+        assertExists(gl.createShader(type), 'Failed to create WebGL shader');
     gl.shaderSource(shader, source);
     gl.compileShader(shader);
     if (!gl.getShaderParameter(shader, gl.COMPILE_STATUS)) {
@@ -119,7 +119,7 @@ export class MPImageShaderContext {
   protected setupShaders(): void {
     const gl = this.gl!;
     this.program =
-        assertNotNull(gl.createProgram()!, 'Failed to create WebGL program');
+        assertExists(gl.createProgram()!, 'Failed to create WebGL program');
 
     this.vertexShader =
         this.compileShader(this.getVertexShader(), gl.VERTEX_SHADER);
@@ -144,11 +144,11 @@ export class MPImageShaderContext {
   private createBuffers(flipVertically: boolean): MPImageShaderBuffers {
     const gl = this.gl!;
     const vertexArrayObject =
-        assertNotNull(gl.createVertexArray(), 'Failed to create vertex array');
+        assertExists(gl.createVertexArray(), 'Failed to create vertex array');
     gl.bindVertexArray(vertexArrayObject);
 
     const vertexBuffer =
-        assertNotNull(gl.createBuffer(), 'Failed to create buffer');
+        assertExists(gl.createBuffer(), 'Failed to create buffer');
     gl.bindBuffer(gl.ARRAY_BUFFER, vertexBuffer);
     gl.enableVertexAttribArray(this.aVertex!);
     gl.vertexAttribPointer(this.aVertex!, 2, gl.FLOAT, false, 0, 0);
@@ -157,7 +157,7 @@ export class MPImageShaderContext {
         gl.STATIC_DRAW);
 
     const textureBuffer =
-        assertNotNull(gl.createBuffer(), 'Failed to create buffer');
+        assertExists(gl.createBuffer(), 'Failed to create buffer');
     gl.bindBuffer(gl.ARRAY_BUFFER, textureBuffer);
     gl.enableVertexAttribArray(this.aTex!);
     gl.vertexAttribPointer(this.aTex!, 2, gl.FLOAT, false, 0, 0);
@@ -232,7 +232,7 @@ export class MPImageShaderContext {
       WebGLTexture {
     this.maybeInitGL(gl);
     const texture =
-        assertNotNull(gl.createTexture(), 'Failed to create texture');
+        assertExists(gl.createTexture(), 'Failed to create texture');
     gl.bindTexture(gl.TEXTURE_2D, texture);
     gl.texParameteri(
         gl.TEXTURE_2D, gl.TEXTURE_WRAP_S, wrapping ?? gl.CLAMP_TO_EDGE);
@@ -252,7 +252,7 @@ export class MPImageShaderContext {
     this.maybeInitGL(gl);
     if (!this.framebuffer) {
       this.framebuffer =
-          assertNotNull(gl.createFramebuffer(), 'Failed to create framebuffe.');
+          assertExists(gl.createFramebuffer(), 'Failed to create framebuffe.');
     }
     gl.bindFramebuffer(gl.FRAMEBUFFER, this.framebuffer);
     gl.framebufferTexture2D(

mediapipe/tasks/web/vision/core/mask.ts

@@ -14,7 +14,7 @@
  * limitations under the License.
  */
 
-import {assertNotNull, MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
+import {assertExists, MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
 import {isIOS} from '../../../../web/graph_runner/platform_utils';
 
 /** Number of instances a user can keep alive before we raise a warning. */
@@ -270,8 +270,8 @@ export class MPMask {
           'is passed when initializing the image.');
     }
     if (!this.gl) {
-      this.gl = assertNotNull(
-          this.canvas.getContext('webgl2'),
+      this.gl = assertExists(
+          this.canvas.getContext('webgl2') as WebGL2RenderingContext,
           'You cannot use a canvas that is already bound to a different ' +
               'type of rendering context.');
     }

mediapipe/tasks/web/vision/core/vision_task_runner.ts

@@ -21,7 +21,7 @@ import {MPImage} from '../../../../tasks/web/vision/core/image';
 import {ImageProcessingOptions} from '../../../../tasks/web/vision/core/image_processing_options';
 import {MPImageShaderContext} from '../../../../tasks/web/vision/core/image_shader_context';
 import {MPMask} from '../../../../tasks/web/vision/core/mask';
-import {GraphRunner, ImageSource, WasmMediaPipeConstructor} from '../../../../web/graph_runner/graph_runner';
+import {getImageSourceSize, GraphRunner, ImageSource, WasmMediaPipeConstructor} from '../../../../web/graph_runner/graph_runner';
 import {SupportImage, WasmImage} from '../../../../web/graph_runner/graph_runner_image_lib';
 import {isWebKit} from '../../../../web/graph_runner/platform_utils';
 import {SupportModelResourcesGraphService} from '../../../../web/graph_runner/register_model_resources_graph_service';
@@ -134,22 +134,6 @@ export abstract class VisionTaskRunner extends TaskRunner {
     this.process(imageFrame, imageProcessingOptions, timestamp);
   }
 
-  private getImageSourceSize(imageSource: ImageSource): [number, number] {
-    if ((imageSource as HTMLVideoElement).videoWidth !== undefined) {
-      return [
-        (imageSource as HTMLVideoElement).videoWidth,
-        (imageSource as HTMLVideoElement).videoHeight
-      ];
-    } else if ((imageSource as HTMLImageElement).naturalWidth !== undefined) {
-      return [
-        (imageSource as HTMLImageElement).naturalWidth,
-        (imageSource as HTMLImageElement).naturalHeight
-      ];
-    } else {
-      return [imageSource.width, imageSource.height];
-    }
-  }
-
   private convertToNormalizedRect(
       imageSource: ImageSource,
       imageProcessingOptions?: ImageProcessingOptions): NormalizedRect {
@@ -199,7 +183,7 @@ export abstract class VisionTaskRunner extends TaskRunner {
       //   uses this for cropping,
       // - then finally rotates this back.
       if (imageProcessingOptions?.rotationDegrees % 180 !== 0) {
-        const [imageWidth, imageHeight] = this.getImageSourceSize(imageSource);
+        const [imageWidth, imageHeight] = getImageSourceSize(imageSource);
         // tslint:disable:no-unnecessary-type-assertion
         const width = normalizedRect.getHeight()! * imageHeight / imageWidth;
         const height = normalizedRect.getWidth()! * imageWidth / imageHeight;

mediapipe/tasks/web/vision/holistic_landmarker/BUILD

@@ -0,0 +1,94 @@
+# This contains the MediaPipe Hand Landmarker Task.
+#
+# This task takes video frames and outputs synchronized frames along with
+# the detection results for one or more holistic categories, using Hand Landmarker.
+
+load("//mediapipe/framework/port:build_config.bzl", "mediapipe_ts_declaration", "mediapipe_ts_library")
+load("@npm//@bazel/jasmine:index.bzl", "jasmine_node_test")
+
+package(default_visibility = ["//mediapipe/tasks:internal"])
+
+licenses(["notice"])
+
+mediapipe_ts_library(
+    name = "holistic_landmarker",
+    srcs = ["holistic_landmarker.ts"],
+    visibility = ["//visibility:public"],
+    deps = [
+        ":holistic_landmarker_connections",
+        ":holistic_landmarker_types",
+        "//mediapipe/framework:calculator_jspb_proto",
+        "//mediapipe/framework:calculator_options_jspb_proto",
+        "//mediapipe/framework/formats:classification_jspb_proto",
+        "//mediapipe/framework/formats:landmark_jspb_proto",
+        "//mediapipe/tasks/cc/core/proto:base_options_jspb_proto",
+        "//mediapipe/tasks/cc/vision/face_detector/proto:face_detector_graph_options_jspb_proto",
+        "//mediapipe/tasks/cc/vision/face_landmarker/proto:face_landmarks_detector_graph_options_jspb_proto",
+        "//mediapipe/tasks/cc/vision/hand_landmarker/proto:hand_landmarks_detector_graph_options_jspb_proto",
+        "//mediapipe/tasks/cc/vision/hand_landmarker/proto:hand_roi_refinement_graph_options_jspb_proto",
+        "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_landmarker_graph_options_jspb_proto",
+        "//mediapipe/tasks/cc/vision/pose_detector/proto:pose_detector_graph_options_jspb_proto",
+        "//mediapipe/tasks/cc/vision/pose_landmarker/proto:pose_landmarks_detector_graph_options_jspb_proto",
+        "//mediapipe/tasks/web/components/containers:category",
+        "//mediapipe/tasks/web/components/containers:classification_result",
+        "//mediapipe/tasks/web/components/containers:landmark",
+        "//mediapipe/tasks/web/components/processors:classifier_result",
+        "//mediapipe/tasks/web/components/processors:landmark_result",
+        "//mediapipe/tasks/web/core",
+        "//mediapipe/tasks/web/vision/core:image_processing_options",
+        "//mediapipe/tasks/web/vision/core:vision_task_runner",
+        "//mediapipe/web/graph_runner:graph_runner_ts",
+    ],
+)
+
+mediapipe_ts_declaration(
+    name = "holistic_landmarker_types",
+    srcs = [
+        "holistic_landmarker_options.d.ts",
+        "holistic_landmarker_result.d.ts",
+    ],
+    visibility = ["//visibility:public"],
+    deps = [
+        "//mediapipe/tasks/web/components/containers:category",
+        "//mediapipe/tasks/web/components/containers:classification_result",
+        "//mediapipe/tasks/web/components/containers:landmark",
+        "//mediapipe/tasks/web/components/containers:matrix",
+        "//mediapipe/tasks/web/core",
+        "//mediapipe/tasks/web/vision/core:mask",
+        "//mediapipe/tasks/web/vision/core:vision_task_options",
+    ],
+)
+
+mediapipe_ts_library(
+    name = "holistic_landmarker_test_lib",
+    testonly = True,
+    srcs = [
+        "holistic_landmarker_test.ts",
+    ],
+    deps = [
+        ":holistic_landmarker",
+        ":holistic_landmarker_types",
+        "//mediapipe/framework:calculator_jspb_proto",
+        "//mediapipe/framework/formats:classification_jspb_proto",
+        "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_landmarker_graph_options_jspb_proto",
+        "//mediapipe/tasks/web/components/processors:landmark_result",
+        "//mediapipe/tasks/web/core",
+        "//mediapipe/tasks/web/core:task_runner_test_utils",
+        "//mediapipe/tasks/web/vision/core:vision_task_runner",
+    ],
+)
+
+jasmine_node_test(
+    name = "holistic_landmarker_test",
+    tags = ["nomsan"],
+    deps = [":holistic_landmarker_test_lib"],
+)
+
+mediapipe_ts_library(
+    name = "holistic_landmarker_connections",
+    deps = [
+        "//mediapipe/tasks/web/vision/face_landmarker:face_landmarks_connections",
+        "//mediapipe/tasks/web/vision/hand_landmarker:hand_landmarks_connections",
+        "//mediapipe/tasks/web/vision/pose_landmarker:pose_landmarks_connections",
+    ],
+)

mediapipe/tasks/web/vision/holistic_landmarker/holistic_landmarker.ts

@@ -0,0 +1,732 @@
+/**
+ * Copyright 2023 The MediaPipe Authors.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+import {Any} from 'google-protobuf/google/protobuf/any_pb';
+import {CalculatorGraphConfig} from '../../../../framework/calculator_pb';
+import {CalculatorOptions} from '../../../../framework/calculator_options_pb';
+import {ClassificationList as ClassificationListProto} from '../../../../framework/formats/classification_pb';
+import {LandmarkList, NormalizedLandmarkList} from '../../../../framework/formats/landmark_pb';
+import {BaseOptions as BaseOptionsProto} from '../../../../tasks/cc/core/proto/base_options_pb';
+import {FaceDetectorGraphOptions} from '../../../../tasks/cc/vision/face_detector/proto/face_detector_graph_options_pb';
+import {FaceLandmarksDetectorGraphOptions} from '../../../../tasks/cc/vision/face_landmarker/proto/face_landmarks_detector_graph_options_pb';
+import {HandLandmarksDetectorGraphOptions} from '../../../../tasks/cc/vision/hand_landmarker/proto/hand_landmarks_detector_graph_options_pb';
+import {HandRoiRefinementGraphOptions} from '../../../../tasks/cc/vision/hand_landmarker/proto/hand_roi_refinement_graph_options_pb';
+import {HolisticLandmarkerGraphOptions} from '../../../../tasks/cc/vision/holistic_landmarker/proto/holistic_landmarker_graph_options_pb';
+import {PoseDetectorGraphOptions} from '../../../../tasks/cc/vision/pose_detector/proto/pose_detector_graph_options_pb';
+import {PoseLandmarksDetectorGraphOptions} from '../../../../tasks/cc/vision/pose_landmarker/proto/pose_landmarks_detector_graph_options_pb';
+import {Classifications} from '../../../../tasks/web/components/containers/classification_result';
+import {Landmark, NormalizedLandmark} from '../../../../tasks/web/components/containers/landmark';
+import {convertFromClassifications} from '../../../../tasks/web/components/processors/classifier_result';
+import {convertToLandmarks, convertToWorldLandmarks} from '../../../../tasks/web/components/processors/landmark_result';
+import {WasmFileset} from '../../../../tasks/web/core/wasm_fileset';
+import {ImageProcessingOptions} from '../../../../tasks/web/vision/core/image_processing_options';
+import {VisionGraphRunner, VisionTaskRunner} from '../../../../tasks/web/vision/core/vision_task_runner';
+import {FACE_LANDMARKS_CONTOURS, FACE_LANDMARKS_FACE_OVAL, FACE_LANDMARKS_LEFT_EYE, FACE_LANDMARKS_LEFT_EYEBROW, FACE_LANDMARKS_LEFT_IRIS, FACE_LANDMARKS_LIPS, FACE_LANDMARKS_RIGHT_EYE, FACE_LANDMARKS_RIGHT_EYEBROW, FACE_LANDMARKS_RIGHT_IRIS, FACE_LANDMARKS_TESSELATION} from '../../../../tasks/web/vision/face_landmarker/face_landmarks_connections';
+import {HAND_CONNECTIONS} from '../../../../tasks/web/vision/hand_landmarker/hand_landmarks_connections';
+import {POSE_CONNECTIONS} from '../../../../tasks/web/vision/pose_landmarker/pose_landmarks_connections';
+import {ImageSource, WasmModule} from '../../../../web/graph_runner/graph_runner';
+// Placeholder for internal dependency on trusted resource url
+
+import {HolisticLandmarkerOptions} from './holistic_landmarker_options';
+import {HolisticLandmarkerResult} from './holistic_landmarker_result';
+
+export * from './holistic_landmarker_options';
+export * from './holistic_landmarker_result';
+export {ImageSource};
+
+// The OSS JS API does not support the builder pattern.
+// tslint:disable:jspb-use-builder-pattern
+
+const IMAGE_STREAM = 'input_frames_image';
+
+const POSE_LANDMARKS_STREAM = 'pose_landmarks';
+const POSE_WORLD_LANDMARKS_STREAM = 'pose_world_landmarks';
+const POSE_SEGMENTATION_MASK_STREAM = 'pose_segmentation_mask';
+const FACE_LANDMARKS_STREAM = 'face_landmarks';
+const FACE_BLENDSHAPES_STREAM = 'extra_blendshapes';
+const LEFT_HAND_LANDMARKS_STREAM = 'left_hand_landmarks';
+const LEFT_HAND_WORLD_LANDMARKS_STREAM = 'left_hand_world_landmarks';
+const RIGHT_HAND_LANDMARKS_STREAM = 'right_hand_landmarks';
+const RIGHT_HAND_WORLD_LANDMARKS_STREAM = 'right_hand_world_landmarks';
+
+const HOLISTIC_LANDMARKER_GRAPH =
+    'mediapipe.tasks.vision.holistic_landmarker.HolisticLandmarkerGraph';
+
+const DEFAULT_SUPRESSION_THRESHOLD = 0.3;
+const DEFAULT_SCORE_THRESHOLD = 0.5;
+
+/**
+ * A callback that receives the result from the holistic landmarker detection.
+ * The returned result are only valid for the duration of the callback. If
+ * asynchronous processing is needed, the masks need to be copied before the
+ * callback returns.
+ */
+export type HolisticLandmarkerCallback = (result: HolisticLandmarkerResult) =>
+    void;
+
+/** Performs holistic landmarks detection on images. */
+export class HolisticLandmarker extends VisionTaskRunner {
+  private result: HolisticLandmarkerResult = {
+    faceLandmarks: [],
+    faceBlendshapes: [],
+    poseLandmarks: [],
+    poseWorldLandmarks: [],
+    poseSegmentationMasks: [],
+    leftHandLandmarks: [],
+    leftHandWorldLandmarks: [],
+    rightHandLandmarks: [],
+    rightHandWorldLandmarks: []
+  };
+  private outputFaceBlendshapes = false;
+  private outputPoseSegmentationMasks = false;
+  private userCallback?: HolisticLandmarkerCallback;
+
+  private readonly options: HolisticLandmarkerGraphOptions;
+  private readonly handLandmarksDetectorGraphOptions:
+      HandLandmarksDetectorGraphOptions;
+  private readonly handRoiRefinementGraphOptions: HandRoiRefinementGraphOptions;
+  private readonly faceDetectorGraphOptions: FaceDetectorGraphOptions;
+  private readonly faceLandmarksDetectorGraphOptions:
+      FaceLandmarksDetectorGraphOptions;
+  private readonly poseDetectorGraphOptions: PoseDetectorGraphOptions;
+  private readonly poseLandmarksDetectorGraphOptions:
+      PoseLandmarksDetectorGraphOptions;
+
+  /**
+   * An array containing the pairs of hand landmark indices to be rendered with
+   * connections.
+   * @export
+   * @nocollapse
+   */
+  static HAND_CONNECTIONS = HAND_CONNECTIONS;
+
+  /**
+   * An array containing the pairs of pose landmark indices to be rendered with
+   * connections.
+   * @export
+   * @nocollapse
+   */
+  static POSE_CONNECTIONS = POSE_CONNECTIONS;
+
+  /**
+   * Landmark connections to draw the connection between a face's lips.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_LIPS = FACE_LANDMARKS_LIPS;
+
+  /**
+   * Landmark connections to draw the connection between a face's left eye.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_LEFT_EYE = FACE_LANDMARKS_LEFT_EYE;
+
+  /**
+   * Landmark connections to draw the connection between a face's left eyebrow.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_LEFT_EYEBROW = FACE_LANDMARKS_LEFT_EYEBROW;
+
+  /**
+   * Landmark connections to draw the connection between a face's left iris.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_LEFT_IRIS = FACE_LANDMARKS_LEFT_IRIS;
+
+  /**
+   * Landmark connections to draw the connection between a face's right eye.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_RIGHT_EYE = FACE_LANDMARKS_RIGHT_EYE;
+
+  /**
+   * Landmark connections to draw the connection between a face's right
+   * eyebrow.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_RIGHT_EYEBROW = FACE_LANDMARKS_RIGHT_EYEBROW;
+
+  /**
+   * Landmark connections to draw the connection between a face's right iris.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_RIGHT_IRIS = FACE_LANDMARKS_RIGHT_IRIS;
+
+  /**
+   * Landmark connections to draw the face's oval.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_FACE_OVAL = FACE_LANDMARKS_FACE_OVAL;
+
+  /**
+   * Landmark connections to draw the face's contour.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_CONTOURS = FACE_LANDMARKS_CONTOURS;
+
+  /**
+   * Landmark connections to draw the face's tesselation.
+   * @export
+   * @nocollapse
+   */
+  static FACE_LANDMARKS_TESSELATION = FACE_LANDMARKS_TESSELATION;
+
+  /**
+   * Initializes the Wasm runtime and creates a new `HolisticLandmarker` from
+   * the provided options.
+   * @export
+   * @param wasmFileset A configuration object that provides the location of the
+   *     Wasm binary and its loader.
+   * @param holisticLandmarkerOptions The options for the HolisticLandmarker.
+   *     Note that either a path to the model asset or a model buffer needs to
+   *     be provided (via `baseOptions`).
+   */
+  static createFromOptions(
+      wasmFileset: WasmFileset,
+      holisticLandmarkerOptions: HolisticLandmarkerOptions):
+      Promise<HolisticLandmarker> {
+    return VisionTaskRunner.createVisionInstance(
+        HolisticLandmarker, wasmFileset, holisticLandmarkerOptions);
+  }
+
+  /**
+   * Initializes the Wasm runtime and creates a new `HolisticLandmarker` based
+   * on the provided model asset buffer.
+   * @export
+   * @param wasmFileset A configuration object that provides the location of the
+   *     Wasm binary and its loader.
+   * @param modelAssetBuffer A binary representation of the model.
+   */
+  static createFromModelBuffer(
+      wasmFileset: WasmFileset,
+      modelAssetBuffer: Uint8Array): Promise<HolisticLandmarker> {
+    return VisionTaskRunner.createVisionInstance(
+        HolisticLandmarker, wasmFileset, {baseOptions: {modelAssetBuffer}});
+  }
+
+  /**
+   * Initializes the Wasm runtime and creates a new `HolisticLandmarker` based
+   * on the path to the model asset.
+   * @export
+   * @param wasmFileset A configuration object that provides the location of the
+   *     Wasm binary and its loader.
+   * @param modelAssetPath The path to the model asset.
+   */
+  static createFromModelPath(
+      wasmFileset: WasmFileset,
+      modelAssetPath: string): Promise<HolisticLandmarker> {
+    return VisionTaskRunner.createVisionInstance(
+        HolisticLandmarker, wasmFileset, {baseOptions: {modelAssetPath}});
+  }
+
+  /** @hideconstructor */
+  constructor(
+      wasmModule: WasmModule,
+      glCanvas?: HTMLCanvasElement|OffscreenCanvas|null) {
+    super(
+        new VisionGraphRunner(wasmModule, glCanvas), IMAGE_STREAM,
+        /* normRectStream= */ null, /* roiAllowed= */ false);
+
+    this.options = new HolisticLandmarkerGraphOptions();
+    this.options.setBaseOptions(new BaseOptionsProto());
+    this.handLandmarksDetectorGraphOptions =
+        new HandLandmarksDetectorGraphOptions();
+    this.options.setHandLandmarksDetectorGraphOptions(
+        this.handLandmarksDetectorGraphOptions);
+    this.handRoiRefinementGraphOptions = new HandRoiRefinementGraphOptions();
+    this.options.setHandRoiRefinementGraphOptions(
+        this.handRoiRefinementGraphOptions);
+    this.faceDetectorGraphOptions = new FaceDetectorGraphOptions();
+    this.options.setFaceDetectorGraphOptions(this.faceDetectorGraphOptions);
+    this.faceLandmarksDetectorGraphOptions =
+        new FaceLandmarksDetectorGraphOptions();
+    this.options.setFaceLandmarksDetectorGraphOptions(
+        this.faceLandmarksDetectorGraphOptions);
+    this.poseDetectorGraphOptions = new PoseDetectorGraphOptions();
+    this.options.setPoseDetectorGraphOptions(this.poseDetectorGraphOptions);
+    this.poseLandmarksDetectorGraphOptions =
+        new PoseLandmarksDetectorGraphOptions();
+    this.options.setPoseLandmarksDetectorGraphOptions(
+        this.poseLandmarksDetectorGraphOptions);
+
+    this.initDefaults();
+  }
+
+  protected override get baseOptions(): BaseOptionsProto {
+    return this.options.getBaseOptions()!;
+  }
+
+  protected override set baseOptions(proto: BaseOptionsProto) {
+    this.options.setBaseOptions(proto);
+  }
+
+  /**
+   * Sets new options for this `HolisticLandmarker`.
+   *
+   * Calling `setOptions()` with a subset of options only affects those options.
+   * You can reset an option back to its default value by explicitly setting it
+   * to `undefined`.
+   *
+   * @export
+   * @param options The options for the holistic landmarker.
+   */
+  override setOptions(options: HolisticLandmarkerOptions): Promise<void> {
+    // Configure face detector options.
+    if ('minFaceDetectionConfidence' in options) {
+      this.faceDetectorGraphOptions.setMinDetectionConfidence(
+          options.minFaceDetectionConfidence ?? DEFAULT_SCORE_THRESHOLD);
+    }
+    if ('minFaceSuppressionThreshold' in options) {
+      this.faceDetectorGraphOptions.setMinSuppressionThreshold(
+          options.minFaceSuppressionThreshold ?? DEFAULT_SUPRESSION_THRESHOLD);
+    }
+
+    // Configure face landmark detector options.
+    if ('minFacePresenceConfidence' in options) {
+      this.faceLandmarksDetectorGraphOptions.setMinDetectionConfidence(
+          options.minFacePresenceConfidence ?? DEFAULT_SCORE_THRESHOLD);
+    }
+    if ('outputFaceBlendshapes' in options) {
+      this.outputFaceBlendshapes = !!options.outputFaceBlendshapes;
+    }
+
+    // Configure pose detector options.
+    if ('minPoseDetectionConfidence' in options) {
+      this.poseDetectorGraphOptions.setMinDetectionConfidence(
+          options.minPoseDetectionConfidence ?? DEFAULT_SCORE_THRESHOLD);
+    }
+    if ('minPoseSuppressionThreshold' in options) {
+      this.poseDetectorGraphOptions.setMinSuppressionThreshold(
+          options.minPoseSuppressionThreshold ?? DEFAULT_SUPRESSION_THRESHOLD);
+    }
+
+    // Configure pose landmark detector options.
+    if ('minPosePresenceConfidence' in options) {
+      this.poseLandmarksDetectorGraphOptions.setMinDetectionConfidence(
+          options.minPosePresenceConfidence ?? DEFAULT_SCORE_THRESHOLD);
+    }
+    if ('outputPoseSegmentationMasks' in options) {
+      this.outputPoseSegmentationMasks = !!options.outputPoseSegmentationMasks;
+    }
+
+    // Configure hand detector options.
+    if ('minHandLandmarksConfidence' in options) {
+      this.handLandmarksDetectorGraphOptions.setMinDetectionConfidence(
+          options.minHandLandmarksConfidence ?? DEFAULT_SCORE_THRESHOLD);
+    }
+    return this.applyOptions(options);
+  }
+
+  /**
+   * Performs holistic landmarks detection on the provided single image and
+   * invokes the callback with the response. The method returns synchronously
+   * once the callback returns. Only use this method when the HolisticLandmarker
+   * is created with running mode `image`.
+   *
+   * @export
+   * @param image An image to process.
+   * @param callback The callback that is invoked with the result. The
+   *    lifetime of the returned masks is only guaranteed for the duration of
+   *    the callback.
+   */
+  detect(image: ImageSource, callback: HolisticLandmarkerCallback): void;
+  /**
+   * Performs holistic landmarks detection on the provided single image and
+   * invokes the callback with the response. The method returns synchronously
+   * once the callback returns. Only use this method when the HolisticLandmarker
+   * is created with running mode `image`.
+   *
+   * @export
+   * @param image An image to process.
+   * @param imageProcessingOptions the `ImageProcessingOptions` specifying how
+   *    to process the input image before running inference.
+   * @param callback The callback that is invoked with the result. The
+   *    lifetime of the returned masks is only guaranteed for the duration of
+   *    the callback.
+   */
+  detect(
+      image: ImageSource, imageProcessingOptions: ImageProcessingOptions,
+      callback: HolisticLandmarkerCallback): void;
+  /**
+   * Performs holistic landmarks detection on the provided single image and
+   * waits synchronously for the response. This method creates a copy of the
+   * resulting masks and should not be used in high-throughput applications.
+   * Only use this method when the HolisticLandmarker is created with running
+   * mode `image`.
+   *
+   * @export
+   * @param image An image to process.
+   * @return The landmarker result. Any masks are copied to avoid lifetime
+   *     limits.
+   * @return The detected pose landmarks.
+   */
+  detect(image: ImageSource): HolisticLandmarkerResult;
+  /**
+   * Performs holistic landmarks detection on the provided single image and
+   * waits synchronously for the response. This method creates a copy of the
+   * resulting masks and should not be used in high-throughput applications.
+   * Only use this method when the HolisticLandmarker is created with running
+   * mode `image`.
+   *
+   * @export
+   * @param image An image to process.
+   * @return The landmarker result. Any masks are copied to avoid lifetime
+   *     limits.
+   * @return The detected pose landmarks.
+   */
+  detect(image: ImageSource, imageProcessingOptions: ImageProcessingOptions):
+      HolisticLandmarkerResult;
+  /** @export */
+  detect(
+      image: ImageSource,
+      imageProcessingOptionsOrCallback?: ImageProcessingOptions|
+      HolisticLandmarkerCallback,
+      callback?: HolisticLandmarkerCallback): HolisticLandmarkerResult|void {
+    const imageProcessingOptions =
+        typeof imageProcessingOptionsOrCallback !== 'function' ?
+        imageProcessingOptionsOrCallback :
+        {};
+    this.userCallback = typeof imageProcessingOptionsOrCallback === 'function' ?
+        imageProcessingOptionsOrCallback :
+        callback!;
+
+    this.resetResults();
+    this.processImageData(image, imageProcessingOptions);
+    return this.processResults();
+  }
+
+  /**
+   * Performs holistic landmarks detection on the provided video frame and
+   * invokes the callback with the response. The method returns synchronously
+   * once the callback returns. Only use this method when the HolisticLandmarker
+   * is created with running mode `video`.
+   *
+   * @export
+   * @param videoFrame A video frame to process.
+   * @param timestamp The timestamp of the current frame, in ms.
+   * @param callback The callback that is invoked with the result. The
+   *    lifetime of the returned masks is only guaranteed for the duration of
+   *    the callback.
+   */
+  detectForVideo(
+      videoFrame: ImageSource, timestamp: number,
+      callback: HolisticLandmarkerCallback): void;
+  /**
+   * Performs holistic landmarks detection on the provided video frame and
+   * invokes the callback with the response. The method returns synchronously
+   * once the callback returns. Only use this method when the holisticLandmarker
+   * is created with running mode `video`.
+   *
+   * @export
+   * @param videoFrame A video frame to process.
+   * @param timestamp The timestamp of the current frame, in ms.
+   * @param imageProcessingOptions the `ImageProcessingOptions` specifying how
+   *    to process the input image before running inference.
+   * @param callback The callback that is invoked with the result. The
+   *    lifetime of the returned masks is only guaranteed for the duration of
+   *    the callback.
+   */
+  detectForVideo(
+      videoFrame: ImageSource, timestamp: number,
+      imageProcessingOptions: ImageProcessingOptions,
+      callback: HolisticLandmarkerCallback): void;
+  /**
+   * Performs holistic landmarks detection on the provided video frame and
+   * returns the result. This method creates a copy of the resulting masks and
+   * should not be used in high-throughput applications. Only use this method
+   * when the HolisticLandmarker is created with running mode `video`.
+   *
+   * @export
+   * @param videoFrame A video frame to process.
+   * @param timestamp The timestamp of the current frame, in ms.
+   * @return The landmarker result. Any masks are copied to extend the
+   *     lifetime of the returned data.
+   */
+  detectForVideo(videoFrame: ImageSource, timestamp: number):
+      HolisticLandmarkerResult;
+  /**
+   * Performs holistic landmarks detection on the provided video frame and waits
+   * synchronously for the response. Only use this method when the
+   * HolisticLandmarker is created with running mode `video`.
+   *
+   * @export
+   * @param videoFrame A video frame to process.
+   * @param timestamp The timestamp of the current frame, in ms.
+   * @param imageProcessingOptions the `ImageProcessingOptions` specifying how
+   *    to process the input image before running inference.
+   * @return The detected holistic landmarks.
+   */
+  detectForVideo(
+      videoFrame: ImageSource, timestamp: number,
+      imageProcessingOptions: ImageProcessingOptions): HolisticLandmarkerResult;
+  /** @export */
+  detectForVideo(
+      videoFrame: ImageSource, timestamp: number,
+      imageProcessingOptionsOrCallback?: ImageProcessingOptions|
+      HolisticLandmarkerCallback,
+      callback?: HolisticLandmarkerCallback): HolisticLandmarkerResult|void {
+    const imageProcessingOptions =
+        typeof imageProcessingOptionsOrCallback !== 'function' ?
+        imageProcessingOptionsOrCallback :
+        {};
+    this.userCallback = typeof imageProcessingOptionsOrCallback === 'function' ?
+        imageProcessingOptionsOrCallback :
+        callback;
+
+    this.resetResults();
+    this.processVideoData(videoFrame, imageProcessingOptions, timestamp);
+    return this.processResults();
+  }
+
+  private resetResults(): void {
+    this.result = {
+      faceLandmarks: [],
+      faceBlendshapes: [],
+      poseLandmarks: [],
+      poseWorldLandmarks: [],
+      poseSegmentationMasks: [],
+      leftHandLandmarks: [],
+      leftHandWorldLandmarks: [],
+      rightHandLandmarks: [],
+      rightHandWorldLandmarks: []
+    };
+  }
+
+  private processResults(): HolisticLandmarkerResult|void {
+    try {
+      if (this.userCallback) {
+        this.userCallback(this.result);
+      } else {
+        return this.result;
+      }
+    } finally {
+      // Free the image memory, now that we've finished our callback.
+      this.freeKeepaliveStreams();
+    }
+  }
+
+  /** Sets the default values for the graph. */
+  private initDefaults(): void {
+    this.faceDetectorGraphOptions.setMinDetectionConfidence(
+        DEFAULT_SCORE_THRESHOLD);
+    this.faceDetectorGraphOptions.setMinSuppressionThreshold(
+        DEFAULT_SUPRESSION_THRESHOLD);
+
+    this.faceLandmarksDetectorGraphOptions.setMinDetectionConfidence(
+        DEFAULT_SCORE_THRESHOLD);
+
+    this.poseDetectorGraphOptions.setMinDetectionConfidence(
+        DEFAULT_SCORE_THRESHOLD);
+    this.poseDetectorGraphOptions.setMinSuppressionThreshold(
+        DEFAULT_SUPRESSION_THRESHOLD);
+
+    this.poseLandmarksDetectorGraphOptions.setMinDetectionConfidence(
+        DEFAULT_SCORE_THRESHOLD);
+
+    this.handLandmarksDetectorGraphOptions.setMinDetectionConfidence(
+        DEFAULT_SCORE_THRESHOLD);
+  }
+
+  /** Converts raw data into a landmark, and adds it to our landmarks list. */
+  private addJsLandmarks(data: Uint8Array, outputList: NormalizedLandmark[][]):
+      void {
+    const landmarksProto = NormalizedLandmarkList.deserializeBinary(data);
+    outputList.push(convertToLandmarks(landmarksProto));
+  }
+
+  /**
+   * Converts raw data into a world landmark, and adds it to our worldLandmarks
+   * list.
+   */
+  private addJsWorldLandmarks(data: Uint8Array, outputList: Landmark[][]):
+      void {
+    const worldLandmarksProto = LandmarkList.deserializeBinary(data);
+    outputList.push(convertToWorldLandmarks(worldLandmarksProto));
+  }
+
+  /** Adds new blendshapes from the given proto. */
+  private addBlenshape(data: Uint8Array, outputList: Classifications[]): void {
+    if (!this.outputFaceBlendshapes) {
+      return;
+    }
+    const classificationList = ClassificationListProto.deserializeBinary(data);
+    outputList.push(convertFromClassifications(
+        classificationList.getClassificationList() ?? []));
+  }
+
+  /** Updates the MediaPipe graph configuration. */
+  protected override refreshGraph(): void {
+    const graphConfig = new CalculatorGraphConfig();
+
+    graphConfig.addInputStream(IMAGE_STREAM);
+    graphConfig.addOutputStream(POSE_LANDMARKS_STREAM);
+    graphConfig.addOutputStream(POSE_WORLD_LANDMARKS_STREAM);
+    graphConfig.addOutputStream(FACE_LANDMARKS_STREAM);
+    graphConfig.addOutputStream(LEFT_HAND_LANDMARKS_STREAM);
+    graphConfig.addOutputStream(LEFT_HAND_WORLD_LANDMARKS_STREAM);
+    graphConfig.addOutputStream(RIGHT_HAND_LANDMARKS_STREAM);
+    graphConfig.addOutputStream(RIGHT_HAND_WORLD_LANDMARKS_STREAM);
+
+    const calculatorOptions = new CalculatorOptions();
+    const optionsProto = new Any();
+    optionsProto.setTypeUrl(
+        'type.googleapis.com/mediapipe.tasks.vision.holistic_landmarker.proto.HolisticLandmarkerGraphOptions');
+    optionsProto.setValue(this.options.serializeBinary());
+
+    const landmarkerNode = new CalculatorGraphConfig.Node();
+    landmarkerNode.setCalculator(HOLISTIC_LANDMARKER_GRAPH);
+    landmarkerNode.addNodeOptions(optionsProto);
+
+    landmarkerNode.addInputStream('IMAGE:' + IMAGE_STREAM);
+    landmarkerNode.addOutputStream('POSE_LANDMARKS:' + POSE_LANDMARKS_STREAM);
+    landmarkerNode.addOutputStream(
+        'POSE_WORLD_LANDMARKS:' + POSE_WORLD_LANDMARKS_STREAM);
+    landmarkerNode.addOutputStream('FACE_LANDMARKS:' + FACE_LANDMARKS_STREAM);
+    landmarkerNode.addOutputStream(
+        'LEFT_HAND_LANDMARKS:' + LEFT_HAND_LANDMARKS_STREAM);
+    landmarkerNode.addOutputStream(
+        'LEFT_HAND_WORLD_LANDMARKS:' + LEFT_HAND_WORLD_LANDMARKS_STREAM);
+    landmarkerNode.addOutputStream(
+        'RIGHT_HAND_LANDMARKS:' + RIGHT_HAND_LANDMARKS_STREAM);
+    landmarkerNode.addOutputStream(
+        'RIGHT_HAND_WORLD_LANDMARKS:' + RIGHT_HAND_WORLD_LANDMARKS_STREAM);
+    landmarkerNode.setOptions(calculatorOptions);
+
+    graphConfig.addNode(landmarkerNode);
+    // We only need to keep alive the image stream, since the protos are being
+    // deep-copied anyways via serialization+deserialization.
+    this.addKeepaliveNode(graphConfig);
+
+    this.graphRunner.attachProtoListener(
+        POSE_LANDMARKS_STREAM, (binaryProto, timestamp) => {
+          this.addJsLandmarks(binaryProto, this.result.poseLandmarks);
+          this.setLatestOutputTimestamp(timestamp);
+        });
+    this.graphRunner.attachEmptyPacketListener(
+        POSE_LANDMARKS_STREAM, timestamp => {
+          this.setLatestOutputTimestamp(timestamp);
+        });
+
+    this.graphRunner.attachProtoListener(
+        POSE_WORLD_LANDMARKS_STREAM, (binaryProto, timestamp) => {
+          this.addJsWorldLandmarks(binaryProto, this.result.poseWorldLandmarks);
+          this.setLatestOutputTimestamp(timestamp);
+        });
+    this.graphRunner.attachEmptyPacketListener(
+        POSE_WORLD_LANDMARKS_STREAM, timestamp => {
+          this.setLatestOutputTimestamp(timestamp);
+        });
+
+    if (this.outputPoseSegmentationMasks) {
+      landmarkerNode.addOutputStream(
+          'POSE_SEGMENTATION_MASK:' + POSE_SEGMENTATION_MASK_STREAM);
+      this.keepStreamAlive(POSE_SEGMENTATION_MASK_STREAM);
+
+      this.graphRunner.attachImageListener(
+          POSE_SEGMENTATION_MASK_STREAM, (mask, timestamp) => {
+            this.result.poseSegmentationMasks = [this.convertToMPMask(
+                mask, /* interpolateValues= */ true,
+                /* shouldCopyData= */ !this.userCallback)];
+            this.setLatestOutputTimestamp(timestamp);
+          });
+      this.graphRunner.attachEmptyPacketListener(
+          POSE_SEGMENTATION_MASK_STREAM, timestamp => {
+            this.result.poseSegmentationMasks = [];
+            this.setLatestOutputTimestamp(timestamp);
+          });
+    }
+
+    this.graphRunner.attachProtoListener(
+        FACE_LANDMARKS_STREAM, (binaryProto, timestamp) => {
+          this.addJsLandmarks(binaryProto, this.result.faceLandmarks);
+          this.setLatestOutputTimestamp(timestamp);
+        });
+    this.graphRunner.attachEmptyPacketListener(
+        FACE_LANDMARKS_STREAM, timestamp => {
+          this.setLatestOutputTimestamp(timestamp);
+        });
+
+    if (this.outputFaceBlendshapes) {
+      graphConfig.addOutputStream(FACE_BLENDSHAPES_STREAM);
+      landmarkerNode.addOutputStream(
+          'FACE_BLENDSHAPES:' + FACE_BLENDSHAPES_STREAM);
+      this.graphRunner.attachProtoListener(
+          FACE_BLENDSHAPES_STREAM, (binaryProto, timestamp) => {
+            this.addBlenshape(binaryProto, this.result.faceBlendshapes);
+            this.setLatestOutputTimestamp(timestamp);
+          });
+      this.graphRunner.attachEmptyPacketListener(
+          FACE_BLENDSHAPES_STREAM, timestamp => {
+            this.setLatestOutputTimestamp(timestamp);
+          });
+    }
+
+    this.graphRunner.attachProtoListener(
+        LEFT_HAND_LANDMARKS_STREAM, (binaryProto, timestamp) => {
+          this.addJsLandmarks(binaryProto, this.result.leftHandLandmarks);
+          this.setLatestOutputTimestamp(timestamp);
+        });
+    this.graphRunner.attachEmptyPacketListener(
+        LEFT_HAND_LANDMARKS_STREAM, timestamp => {
+          this.setLatestOutputTimestamp(timestamp);
+        });
+
+    this.graphRunner.attachProtoListener(
+        LEFT_HAND_WORLD_LANDMARKS_STREAM, (binaryProto, timestamp) => {
+          this.addJsWorldLandmarks(
+              binaryProto, this.result.leftHandWorldLandmarks);
+          this.setLatestOutputTimestamp(timestamp);
+        });
+    this.graphRunner.attachEmptyPacketListener(
+        LEFT_HAND_WORLD_LANDMARKS_STREAM, timestamp => {
+          this.setLatestOutputTimestamp(timestamp);
+        });
+
+    this.graphRunner.attachProtoListener(
+        RIGHT_HAND_LANDMARKS_STREAM, (binaryProto, timestamp) => {
+          this.addJsLandmarks(binaryProto, this.result.rightHandLandmarks);
+          this.setLatestOutputTimestamp(timestamp);
+        });
+    this.graphRunner.attachEmptyPacketListener(
+        RIGHT_HAND_LANDMARKS_STREAM, timestamp => {
+          this.setLatestOutputTimestamp(timestamp);
+        });
+
+    this.graphRunner.attachProtoListener(
+        RIGHT_HAND_WORLD_LANDMARKS_STREAM, (binaryProto, timestamp) => {
+          this.addJsWorldLandmarks(
+              binaryProto, this.result.rightHandWorldLandmarks);
+          this.setLatestOutputTimestamp(timestamp);
+        });
+    this.graphRunner.attachEmptyPacketListener(
+        RIGHT_HAND_WORLD_LANDMARKS_STREAM, timestamp => {
+          this.setLatestOutputTimestamp(timestamp);
+        });
+
+    const binaryGraph = graphConfig.serializeBinary();
+    this.setGraph(new Uint8Array(binaryGraph), /* isBinary= */ true);
+  }
+}
+
+

mediapipe/tasks/web/vision/holistic_landmarker/holistic_landmarker_options.d.ts

@@ -0,0 +1,71 @@
+/**
+ * Copyright 2023 The MediaPipe Authors.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+import {VisionTaskOptions} from '../../../../tasks/web/vision/core/vision_task_options';
+
+/** Options to configure the MediaPipe HolisticLandmarker Task */
+export declare interface HolisticLandmarkerOptions extends VisionTaskOptions {
+  /**
+   * The minimum confidence score for the face detection to be considered
+   * successful. Defaults to 0.5.
+   */
+  minFaceDetectionConfidence?: number|undefined;
+
+  /**
+   * The minimum non-maximum-suppression threshold for face detection to be
+   * considered overlapped. Defaults to 0.3.
+   */
+  minFaceSuppressionThreshold?: number|undefined;
+
+  /**
+   * The minimum confidence score of face presence score in the face landmarks
+   * detection. Defaults to 0.5.
+   */
+  minFacePresenceConfidence?: number|undefined;
+
+  /**
+   * Whether FaceLandmarker outputs face blendshapes classification. Face
+   * blendshapes are used for rendering the 3D face model.
+   */
+  outputFaceBlendshapes?: boolean|undefined;
+
+  /**
+   * The minimum confidence score for the pose detection to be considered
+   * successful. Defaults to 0.5.
+   */
+  minPoseDetectionConfidence?: number|undefined;
+
+  /**
+   * The minimum non-maximum-suppression threshold for pose detection to be
+   * considered overlapped. Defaults to 0.3.
+   */
+  minPoseSuppressionThreshold?: number|undefined;
+
+  /**
+   * The minimum confidence score of pose presence score in the pose landmarks
+   * detection. Defaults to 0.5.
+   */
+  minPosePresenceConfidence?: number|undefined;
+
+  /** Whether to output segmentation masks. Defaults to false. */
+  outputPoseSegmentationMasks?: boolean|undefined;
+
+  /**
+   * The minimum confidence score of hand presence score in the hand landmarks
+   * detection. Defaults to 0.5.
+   */
+  minHandLandmarksConfidence?: number|undefined;
+}

mediapipe/tasks/web/vision/holistic_landmarker/holistic_landmarker_result.d.ts

@@ -0,0 +1,55 @@
+/**
+ * Copyright 2023 The MediaPipe Authors.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+import {Category} from '../../../../tasks/web/components/containers/category';
+import {Classifications} from '../../../../tasks/web/components/containers/classification_result';
+import {Landmark, NormalizedLandmark} from '../../../../tasks/web/components/containers/landmark';
+import {MPMask} from '../../../../tasks/web/vision/core/mask';
+
+export {Category, Landmark, NormalizedLandmark};
+
+/**
+ * Represents the holistic landmarks detection results generated by
+ * `HolisticLandmarker`.
+ */
+export declare interface HolisticLandmarkerResult {
+  /** Detected face landmarks in normalized image coordinates. */
+  faceLandmarks: NormalizedLandmark[][];
+
+  /** Optional face blendshapes results. */
+  faceBlendshapes: Classifications[];
+
+  /** Detected pose landmarks in normalized image coordinates. */
+  poseLandmarks: NormalizedLandmark[][];
+
+  /** Pose landmarks in world coordinates of detected poses. */
+  poseWorldLandmarks: Landmark[][];
+
+  /** Optional segmentation mask for the detected pose. */
+  poseSegmentationMasks: MPMask[];
+
+  /** Left hand landmarks of detected left hands. */
+  leftHandLandmarks: NormalizedLandmark[][];
+
+  /** Left hand landmarks in world coordinates of detected left hands. */
+  leftHandWorldLandmarks: Landmark[][];
+
+  /** Right hand landmarks of detected right hands. */
+  rightHandLandmarks: NormalizedLandmark[][];
+
+  /** Right hand landmarks in world coordinates of detected right hands. */
+  rightHandWorldLandmarks: Landmark[][];
+}

mediapipe/tasks/web/vision/holistic_landmarker/holistic_landmarker_test.ts

@@ -0,0 +1,403 @@
+/**
+ * Copyright 2023 The MediaPipe Authors.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+import 'jasmine';
+
+import {CalculatorGraphConfig} from '../../../../framework/calculator_pb';
+import {Classification, ClassificationList} from '../../../../framework/formats/classification_pb';
+import {HolisticLandmarkerGraphOptions} from '../../../../tasks/cc/vision/holistic_landmarker/proto/holistic_landmarker_graph_options_pb';
+import {createLandmarks, createWorldLandmarks} from '../../../../tasks/web/components/processors/landmark_result_test_lib';
+import {addJasmineCustomFloatEqualityTester, createSpyWasmModule, Deserializer, MediapipeTasksFake, SpyWasmModule, verifyGraph, verifyListenersRegistered} from '../../../../tasks/web/core/task_runner_test_utils';
+import {VisionGraphRunner} from '../../../../tasks/web/vision/core/vision_task_runner';
+
+import {HolisticLandmarker} from './holistic_landmarker';
+import {HolisticLandmarkerOptions} from './holistic_landmarker_options';
+
+
+// The OSS JS API does not support the builder pattern.
+// tslint:disable:jspb-use-builder-pattern
+
+type ProtoListener = ((binaryProtos: Uint8Array, timestamp: number) => void);
+const holisticLandmarkerDeserializer =
+    (binaryProto =>
+         HolisticLandmarkerGraphOptions.deserializeBinary(binaryProto)
+             .toObject()) as Deserializer;
+
+function createBlendshapes(): ClassificationList {
+  const blendshapesProto = new ClassificationList();
+  const classification = new Classification();
+  classification.setScore(0.1);
+  classification.setIndex(1);
+  classification.setLabel('face_label');
+  classification.setDisplayName('face_display_name');
+  blendshapesProto.addClassification(classification);
+  return blendshapesProto;
+}
+
+class HolisticLandmarkerFake extends HolisticLandmarker implements
+    MediapipeTasksFake {
+  calculatorName =
+      'mediapipe.tasks.vision.holistic_landmarker.HolisticLandmarkerGraph';
+  attachListenerSpies: jasmine.Spy[] = [];
+  graph: CalculatorGraphConfig|undefined;
+  fakeWasmModule: SpyWasmModule;
+  listeners = new Map<string, ProtoListener>();
+
+  constructor() {
+    super(createSpyWasmModule(), /* glCanvas= */ null);
+    this.fakeWasmModule =
+        this.graphRunner.wasmModule as unknown as SpyWasmModule;
+
+    this.attachListenerSpies[0] =
+        spyOn(this.graphRunner, 'attachProtoListener')
+            .and.callFake((stream, listener) => {
+              expect(stream).toMatch(
+                  /(pose_landmarks|pose_world_landmarks|pose_segmentation_mask|face_landmarks|extra_blendshapes|left_hand_landmarks|left_hand_world_landmarks|right_hand_landmarks|right_hand_world_landmarks)/);
+              this.listeners.set(stream, listener);
+            });
+
+    spyOn(this.graphRunner, 'setGraph').and.callFake(binaryGraph => {
+      this.graph = CalculatorGraphConfig.deserializeBinary(binaryGraph);
+    });
+    spyOn(this.graphRunner, 'addGpuBufferAsImageToStream');
+    spyOn(this.graphRunner, 'addProtoToStream');
+  }
+
+  getGraphRunner(): VisionGraphRunner {
+    return this.graphRunner;
+  }
+}
+
+describe('HolisticLandmarker', () => {
+  let holisticLandmarker: HolisticLandmarkerFake;
+
+  beforeEach(async () => {
+    addJasmineCustomFloatEqualityTester();
+    holisticLandmarker = new HolisticLandmarkerFake();
+    await holisticLandmarker.setOptions(
+        {baseOptions: {modelAssetBuffer: new Uint8Array([])}});
+  });
+
+  afterEach(() => {
+    holisticLandmarker.close();
+  });
+
+  it('initializes graph', async () => {
+    verifyGraph(holisticLandmarker);
+    verifyGraph(
+        holisticLandmarker, undefined, undefined,
+        holisticLandmarkerDeserializer);
+  });
+
+  it('reloads graph when settings are changed', async () => {
+    verifyListenersRegistered(holisticLandmarker);
+
+    await holisticLandmarker.setOptions({minFaceDetectionConfidence: 0.6});
+    verifyGraph(
+        holisticLandmarker,
+        [['faceDetectorGraphOptions', 'minDetectionConfidence'], 0.6],
+        undefined, holisticLandmarkerDeserializer);
+    verifyListenersRegistered(holisticLandmarker);
+
+    await holisticLandmarker.setOptions({minFaceDetectionConfidence: 0.7});
+    verifyGraph(
+        holisticLandmarker,
+        [['faceDetectorGraphOptions', 'minDetectionConfidence'], 0.7],
+        undefined, holisticLandmarkerDeserializer);
+    verifyListenersRegistered(holisticLandmarker);
+  });
+
+  it('merges options', async () => {
+    await holisticLandmarker.setOptions({minFaceDetectionConfidence: 0.5});
+    await holisticLandmarker.setOptions({minFaceSuppressionThreshold: 0.5});
+    await holisticLandmarker.setOptions({minFacePresenceConfidence: 0.5});
+    await holisticLandmarker.setOptions({minPoseDetectionConfidence: 0.5});
+    await holisticLandmarker.setOptions({minPoseSuppressionThreshold: 0.5});
+    await holisticLandmarker.setOptions({minPosePresenceConfidence: 0.5});
+    await holisticLandmarker.setOptions({minHandLandmarksConfidence: 0.5});
+
+    verifyGraph(
+        holisticLandmarker,
+        [
+          'faceDetectorGraphOptions', {
+            baseOptions: undefined,
+            minDetectionConfidence: 0.5,
+            minSuppressionThreshold: 0.5,
+            numFaces: undefined
+          }
+        ],
+        undefined, holisticLandmarkerDeserializer);
+    verifyGraph(
+        holisticLandmarker,
+        [
+          'faceLandmarksDetectorGraphOptions', {
+            baseOptions: undefined,
+            minDetectionConfidence: 0.5,
+            smoothLandmarks: undefined,
+            faceBlendshapesGraphOptions: undefined
+          }
+        ],
+        undefined, holisticLandmarkerDeserializer);
+    verifyGraph(
+        holisticLandmarker,
+        [
+          'poseDetectorGraphOptions', {
+            baseOptions: undefined,
+            minDetectionConfidence: 0.5,
+            minSuppressionThreshold: 0.5,
+            numPoses: undefined
+          }
+        ],
+        undefined, holisticLandmarkerDeserializer);
+    verifyGraph(
+        holisticLandmarker,
+        [
+          'poseLandmarksDetectorGraphOptions', {
+            baseOptions: undefined,
+            minDetectionConfidence: 0.5,
+            smoothLandmarks: undefined
+          }
+        ],
+        undefined, holisticLandmarkerDeserializer);
+    verifyGraph(
+        holisticLandmarker,
+        [
+          'handLandmarksDetectorGraphOptions',
+          {baseOptions: undefined, minDetectionConfidence: 0.5}
+        ],
+        undefined, holisticLandmarkerDeserializer);
+  });
+
+  describe('setOptions()', () => {
+    interface TestCase {
+      optionPath: [keyof HolisticLandmarkerOptions, ...string[]];
+      fieldPath: string[];
+      customValue: unknown;
+      defaultValue: unknown;
+    }
+
+    const testCases: TestCase[] = [
+      {
+        optionPath: ['minFaceDetectionConfidence'],
+        fieldPath: ['faceDetectorGraphOptions', 'minDetectionConfidence'],
+        customValue: 0.1,
+        defaultValue: 0.5
+      },
+      {
+        optionPath: ['minFaceSuppressionThreshold'],
+        fieldPath: ['faceDetectorGraphOptions', 'minSuppressionThreshold'],
+        customValue: 0.2,
+        defaultValue: 0.3
+      },
+      {
+        optionPath: ['minFacePresenceConfidence'],
+        fieldPath:
+            ['faceLandmarksDetectorGraphOptions', 'minDetectionConfidence'],
+        customValue: 0.2,
+        defaultValue: 0.5
+      },
+      {
+        optionPath: ['minPoseDetectionConfidence'],
+        fieldPath: ['poseDetectorGraphOptions', 'minDetectionConfidence'],
+        customValue: 0.1,
+        defaultValue: 0.5
+      },
+      {
+        optionPath: ['minPoseSuppressionThreshold'],
+        fieldPath: ['poseDetectorGraphOptions', 'minSuppressionThreshold'],
+        customValue: 0.2,
+        defaultValue: 0.3
+      },
+      {
+        optionPath: ['minPosePresenceConfidence'],
+        fieldPath:
+            ['poseLandmarksDetectorGraphOptions', 'minDetectionConfidence'],
+        customValue: 0.2,
+        defaultValue: 0.5
+      },
+      {
+        optionPath: ['minHandLandmarksConfidence'],
+        fieldPath:
+            ['handLandmarksDetectorGraphOptions', 'minDetectionConfidence'],
+        customValue: 0.1,
+        defaultValue: 0.5
+      },
+    ];
+
+    /** Creates an options object that can be passed to setOptions() */
+    function createOptions(
+        path: string[], value: unknown): HolisticLandmarkerOptions {
+      const options: Record<string, unknown> = {};
+      let currentLevel = options;
+      for (const element of path.slice(0, -1)) {
+        currentLevel[element] = {};
+        currentLevel = currentLevel[element] as Record<string, unknown>;
+      }
+      currentLevel[path[path.length - 1]] = value;
+      return options;
+    }
+
+    for (const testCase of testCases) {
+      it(`uses default value for ${testCase.optionPath[0]}`, async () => {
+        verifyGraph(
+            holisticLandmarker, [testCase.fieldPath, testCase.defaultValue],
+            undefined, holisticLandmarkerDeserializer);
+      });
+
+      it(`can set ${testCase.optionPath[0]}`, async () => {
+        await holisticLandmarker.setOptions(
+            createOptions(testCase.optionPath, testCase.customValue));
+        verifyGraph(
+            holisticLandmarker, [testCase.fieldPath, testCase.customValue],
+            undefined, holisticLandmarkerDeserializer);
+      });
+
+      it(`can clear ${testCase.optionPath[0]}`, async () => {
+        await holisticLandmarker.setOptions(
+            createOptions(testCase.optionPath, testCase.customValue));
+        verifyGraph(
+            holisticLandmarker, [testCase.fieldPath, testCase.customValue],
+            undefined, holisticLandmarkerDeserializer);
+
+        await holisticLandmarker.setOptions(
+            createOptions(testCase.optionPath, undefined));
+        verifyGraph(
+            holisticLandmarker, [testCase.fieldPath, testCase.defaultValue],
+            undefined, holisticLandmarkerDeserializer);
+      });
+    }
+  });
+
+  it('supports outputFaceBlendshapes', async () => {
+    const stream = 'extra_blendshapes';
+    await holisticLandmarker.setOptions({});
+    expect(holisticLandmarker.graph!.getOutputStreamList())
+        .not.toContain(stream);
+
+    await holisticLandmarker.setOptions({outputFaceBlendshapes: false});
+    expect(holisticLandmarker.graph!.getOutputStreamList())
+        .not.toContain(stream);
+
+    await holisticLandmarker.setOptions({outputFaceBlendshapes: true});
+    expect(holisticLandmarker.graph!.getOutputStreamList()).toContain(stream);
+  });
+
+  it('transforms results', async () => {
+    const faceLandmarksProto = createLandmarks().serializeBinary();
+    const blendshapesProto = createBlendshapes().serializeBinary();
+
+    const poseLandmarksProto = createLandmarks().serializeBinary();
+    const poseWorldLandmarksProto = createWorldLandmarks().serializeBinary();
+
+    const leftHandLandmarksProto = createLandmarks().serializeBinary();
+    const leftHandWorldLandmarksProto =
+        createWorldLandmarks().serializeBinary();
+    const rightHandLandmarksProto = createLandmarks().serializeBinary();
+    const rightHandWorldLandmarksProto =
+        createWorldLandmarks().serializeBinary();
+
+    await holisticLandmarker.setOptions(
+        {outputFaceBlendshapes: true, outputPoseSegmentationMasks: false});
+
+    // Pass the test data to our listener
+    holisticLandmarker.fakeWasmModule._waitUntilIdle.and.callFake(() => {
+      verifyListenersRegistered(holisticLandmarker);
+      holisticLandmarker.listeners.get('face_landmarks')!
+          (faceLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('extra_blendshapes')!
+          (blendshapesProto, 1337);
+
+      holisticLandmarker.listeners.get('pose_landmarks')!
+          (poseLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('pose_world_landmarks')!
+          (poseWorldLandmarksProto, 1337);
+
+      holisticLandmarker.listeners.get('left_hand_landmarks')!
+          (leftHandLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('left_hand_world_landmarks')!
+          (leftHandWorldLandmarksProto, 1337);
+
+      holisticLandmarker.listeners.get('right_hand_landmarks')!
+          (rightHandLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('right_hand_world_landmarks')!
+          (rightHandWorldLandmarksProto, 1337);
+    });
+
+    // Invoke the holistic landmarker
+    const landmarks = holisticLandmarker.detect({} as HTMLImageElement);
+    expect(holisticLandmarker.getGraphRunner().addGpuBufferAsImageToStream)
+        .toHaveBeenCalledTimes(1);
+    expect(holisticLandmarker.fakeWasmModule._waitUntilIdle).toHaveBeenCalled();
+
+    expect(landmarks).toEqual({
+      faceLandmarks: [[{x: 0, y: 0, z: 0}]],
+      faceBlendshapes: [{
+        categories: [{
+          index: 1,
+          score: 0.1,
+          categoryName: 'face_label',
+          displayName: 'face_display_name'
+        }],
+        headIndex: -1,
+        headName: ''
+      }],
+      poseLandmarks: [[{x: 0, y: 0, z: 0}]],
+      poseWorldLandmarks: [[{x: 0, y: 0, z: 0}]],
+      poseSegmentationMasks: [],
+      leftHandLandmarks: [[{x: 0, y: 0, z: 0}]],
+      leftHandWorldLandmarks: [[{x: 0, y: 0, z: 0}]],
+      rightHandLandmarks: [[{x: 0, y: 0, z: 0}]],
+      rightHandWorldLandmarks: [[{x: 0, y: 0, z: 0}]]
+    });
+  });
+
+  it('clears results between invoations', async () => {
+    const faceLandmarksProto = createLandmarks().serializeBinary();
+    const poseLandmarksProto = createLandmarks().serializeBinary();
+    const poseWorldLandmarksProto = createWorldLandmarks().serializeBinary();
+    const leftHandLandmarksProto = createLandmarks().serializeBinary();
+    const leftHandWorldLandmarksProto =
+        createWorldLandmarks().serializeBinary();
+    const rightHandLandmarksProto = createLandmarks().serializeBinary();
+    const rightHandWorldLandmarksProto =
+        createWorldLandmarks().serializeBinary();
+
+    // Pass the test data to our listener
+    holisticLandmarker.fakeWasmModule._waitUntilIdle.and.callFake(() => {
+      holisticLandmarker.listeners.get('face_landmarks')!
+          (faceLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('pose_landmarks')!
+          (poseLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('pose_world_landmarks')!
+          (poseWorldLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('left_hand_landmarks')!
+          (leftHandLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('left_hand_world_landmarks')!
+          (leftHandWorldLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('right_hand_landmarks')!
+          (rightHandLandmarksProto, 1337);
+      holisticLandmarker.listeners.get('right_hand_world_landmarks')!
+          (rightHandWorldLandmarksProto, 1337);
+    });
+
+    // Invoke the holistic landmarker twice
+    const landmarks1 = holisticLandmarker.detect({} as HTMLImageElement);
+    const landmarks2 = holisticLandmarker.detect({} as HTMLImageElement);
+
+    // Verify that landmarks2 is not a concatenation of all previously returned
+    // hands.
+    expect(landmarks1).toEqual(landmarks2);
+  });
+});

mediapipe/tasks/web/vision/index.ts

@@ -23,6 +23,7 @@ import {FaceLandmarker as FaceLandmarkerImpl} from '../../../tasks/web/vision/fa
 import {FaceStylizer as FaceStylizerImpl} from '../../../tasks/web/vision/face_stylizer/face_stylizer';
 import {GestureRecognizer as GestureRecognizerImpl} from '../../../tasks/web/vision/gesture_recognizer/gesture_recognizer';
 import {HandLandmarker as HandLandmarkerImpl} from '../../../tasks/web/vision/hand_landmarker/hand_landmarker';
+import {HolisticLandmarker as HolisticLandmarkerImpl} from '../../../tasks/web/vision/holistic_landmarker/holistic_landmarker';
 import {ImageClassifier as ImageClassifierImpl} from '../../../tasks/web/vision/image_classifier/image_classifier';
 import {ImageEmbedder as ImageEmbedderImpl} from '../../../tasks/web/vision/image_embedder/image_embedder';
 import {ImageSegmenter as ImageSegementerImpl} from '../../../tasks/web/vision/image_segmenter/image_segmenter';
@@ -41,6 +42,7 @@ const FaceLandmarker = FaceLandmarkerImpl;
 const FaceStylizer = FaceStylizerImpl;
 const GestureRecognizer = GestureRecognizerImpl;
 const HandLandmarker = HandLandmarkerImpl;
+const HolisticLandmarker = HolisticLandmarkerImpl;
 const ImageClassifier = ImageClassifierImpl;
 const ImageEmbedder = ImageEmbedderImpl;
 const ImageSegmenter = ImageSegementerImpl;
@@ -58,6 +60,7 @@ export {
   FaceStylizer,
   GestureRecognizer,
   HandLandmarker,
+  HolisticLandmarker,
   ImageClassifier,
   ImageEmbedder,
   ImageSegmenter,

mediapipe/tasks/web/vision/pose_landmarker/BUILD

@@ -19,6 +19,7 @@ mediapipe_ts_library(
     visibility = ["//visibility:public"],
     deps = [
         ":pose_landmarker_types",
+        ":pose_landmarks_connections",
         "//mediapipe/framework:calculator_jspb_proto",
         "//mediapipe/framework:calculator_options_jspb_proto",
         "//mediapipe/framework/formats:landmark_jspb_proto",
@@ -32,7 +33,6 @@ mediapipe_ts_library(
         "//mediapipe/tasks/web/core",
         "//mediapipe/tasks/web/vision/core:image_processing_options",
         "//mediapipe/tasks/web/vision/core:mask",
-        "//mediapipe/tasks/web/vision/core:types",
         "//mediapipe/tasks/web/vision/core:vision_task_runner",
         "//mediapipe/web/graph_runner:graph_runner_ts",
     ],
@@ -67,3 +67,9 @@ jasmine_node_test(
     tags = ["nomsan"],
     deps = [":pose_landmarker_test_lib"],
 )
+
+mediapipe_ts_library(
+    name = "pose_landmarks_connections",
+    srcs = ["pose_landmarks_connections.ts"],
+    deps = ["//mediapipe/tasks/web/vision/core:types"],
+)

mediapipe/tasks/web/vision/pose_landmarker/pose_landmarker.ts

@@ -26,13 +26,13 @@ import {convertToLandmarks, convertToWorldLandmarks} from '../../../../tasks/web
 import {WasmFileset} from '../../../../tasks/web/core/wasm_fileset';
 import {ImageProcessingOptions} from '../../../../tasks/web/vision/core/image_processing_options';
 import {MPMask} from '../../../../tasks/web/vision/core/mask';
-import {convertToConnections} from '../../../../tasks/web/vision/core/types';
 import {VisionGraphRunner, VisionTaskRunner} from '../../../../tasks/web/vision/core/vision_task_runner';
 import {ImageSource, WasmModule} from '../../../../web/graph_runner/graph_runner';
 // Placeholder for internal dependency on trusted resource url
 
 import {PoseLandmarkerOptions} from './pose_landmarker_options';
 import {PoseLandmarkerResult} from './pose_landmarker_result';
+import {POSE_CONNECTIONS} from './pose_landmarks_connections';
 
 export * from './pose_landmarker_options';
 export * from './pose_landmarker_result';
@@ -79,12 +79,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * @export
    * @nocollapse
    */
-  static POSE_CONNECTIONS = convertToConnections(
-      [0, 1], [1, 2], [2, 3], [3, 7], [0, 4], [4, 5], [5, 6], [6, 8], [9, 10],
-      [11, 12], [11, 13], [13, 15], [15, 17], [15, 19], [15, 21], [17, 19],
-      [12, 14], [14, 16], [16, 18], [16, 20], [16, 22], [18, 20], [11, 23],
-      [12, 24], [23, 24], [23, 25], [24, 26], [25, 27], [26, 28], [27, 29],
-      [28, 30], [29, 31], [30, 32], [27, 31], [28, 32]);
+  static POSE_CONNECTIONS = POSE_CONNECTIONS;
 
   /**
    * Initializes the Wasm runtime and creates a new `PoseLandmarker` from the
@@ -206,6 +201,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * callback returns. Only use this method when the PoseLandmarker is created
    * with running mode `image`.
    *
+   * @export
    * @param image An image to process.
    * @param callback The callback that is invoked with the result. The
    *    lifetime of the returned masks is only guaranteed for the duration of
@@ -218,6 +214,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * callback returns. Only use this method when the PoseLandmarker is created
    * with running mode `image`.
    *
+   * @export
    * @param image An image to process.
    * @param imageProcessingOptions the `ImageProcessingOptions` specifying how
    *    to process the input image before running inference.
@@ -235,6 +232,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * use this method when the PoseLandmarker is created with running mode
    * `image`.
    *
+   * @export
    * @param image An image to process.
    * @return The landmarker result. Any masks are copied to avoid lifetime
    *     limits.
@@ -248,6 +246,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * use this method when the PoseLandmarker is created with running mode
    * `image`.
    *
+   * @export
    * @param image An image to process.
    * @return The landmarker result. Any masks are copied to avoid lifetime
    *     limits.
@@ -280,6 +279,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * callback returns. Only use this method when the PoseLandmarker is created
    * with running mode `video`.
    *
+   * @export
    * @param videoFrame A video frame to process.
    * @param timestamp The timestamp of the current frame, in ms.
    * @param callback The callback that is invoked with the result. The
@@ -295,6 +295,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * callback returns. Only use this method when the PoseLandmarker is created
    * with running mode `video`.
    *
+   * @export
    * @param videoFrame A video frame to process.
    * @param timestamp The timestamp of the current frame, in ms.
    * @param imageProcessingOptions the `ImageProcessingOptions` specifying how
@@ -313,6 +314,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * in high-throughput applications. Only use this method when the
    * PoseLandmarker is created with running mode `video`.
    *
+   * @export
    * @param videoFrame A video frame to process.
    * @param timestamp The timestamp of the current frame, in ms.
    * @return The landmarker result. Any masks are copied to extend the
@@ -327,6 +329,7 @@ export class PoseLandmarker extends VisionTaskRunner {
    * callback returns. Only use this method when the PoseLandmarker is created
    * with running mode `video`.
    *
+   * @export
    * @param videoFrame A video frame to process.
    * @param timestamp The timestamp of the current frame, in ms.
    * @param imageProcessingOptions the `ImageProcessingOptions` specifying how

mediapipe/tasks/web/vision/pose_landmarker/pose_landmarks_connections.ts

@@ -0,0 +1,28 @@
+/**
+ * Copyright 2023 The MediaPipe Authors.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+import {convertToConnections} from '../../../../tasks/web/vision/core/types';
+
+/**
+ * An array containing the pairs of pose landmark indices to be rendered with
+ * connections.
+ */
+export const POSE_CONNECTIONS = convertToConnections(
+    [0, 1], [1, 2], [2, 3], [3, 7], [0, 4], [4, 5], [5, 6], [6, 8], [9, 10],
+    [11, 12], [11, 13], [13, 15], [15, 17], [15, 19], [15, 21], [17, 19],
+    [12, 14], [14, 16], [16, 18], [16, 20], [16, 22], [18, 20], [11, 23],
+    [12, 24], [23, 24], [23, 25], [24, 26], [25, 27], [26, 28], [27, 29],
+    [28, 30], [29, 31], [30, 32], [27, 31], [28, 32]);

mediapipe/tasks/web/vision/types.ts

@@ -22,6 +22,7 @@ export * from '../../../tasks/web/vision/face_detector/face_detector';
 export * from '../../../tasks/web/vision/face_landmarker/face_landmarker';
 export * from '../../../tasks/web/vision/face_stylizer/face_stylizer';
 export * from '../../../tasks/web/vision/gesture_recognizer/gesture_recognizer';
+export * from '../../../tasks/web/vision/holistic_landmarker/holistic_landmarker';
 export * from '../../../tasks/web/vision/hand_landmarker/hand_landmarker';
 export * from '../../../tasks/web/vision/image_classifier/image_classifier';
 export * from '../../../tasks/web/vision/image_embedder/image_embedder';

mediapipe/util/audio_decoder.h

@@ -90,7 +90,7 @@ class BasePacketProcessor {
   // CorrectPtsForRollover(2^33 - 1) -> 2^33 - 1
   // CorrectPtsForRollover(0) -> 2^33  // PTS in media rolls over, corrected.
   // CorrectPtsForRollover(1) -> 2^33 + 1
-  int64 CorrectPtsForRollover(int64 media_pts);
+  int64_t CorrectPtsForRollover(int64_t media_pts);
 
   AVCodecContext* avcodec_ctx_ = nullptr;
   const AVCodec* avcodec_ = nullptr;
@@ -113,7 +113,7 @@ class BasePacketProcessor {
   AVRational source_frame_rate_;
 
   // The number of frames that were successfully processed.
-  int64 num_frames_processed_ = 0;
+  int64_t num_frames_processed_ = 0;
 
   int bytes_per_sample_ = 0;
 
@@ -121,7 +121,7 @@ class BasePacketProcessor {
   bool last_frame_time_regression_detected_ = false;
 
   // The last rollover corrected PTS returned by CorrectPtsForRollover.
-  int64 rollover_corrected_last_pts_ = AV_NOPTS_VALUE;
+  int64_t rollover_corrected_last_pts_ = AV_NOPTS_VALUE;
 
   // The buffer of current frames.
   std::deque<Packet> buffer_;
@@ -141,16 +141,16 @@ class AudioPacketProcessor : public BasePacketProcessor {
  private:
   // Appends audio in buffer(s) to the output buffer (buffer_).
   absl::Status AddAudioDataToBuffer(const Timestamp output_timestamp,
-                                    uint8* const* raw_audio,
+                                    uint8_t* const* raw_audio,
                                     int buf_size_bytes);
 
   // Converts a number of samples into an approximate stream timestamp value.
-  int64 SampleNumberToTimestamp(const int64 sample_number);
-  int64 TimestampToSampleNumber(const int64 timestamp);
+  int64_t SampleNumberToTimestamp(const int64_t sample_number);
+  int64_t TimestampToSampleNumber(const int64_t timestamp);
 
   // Converts a timestamp/sample number to microseconds.
-  int64 TimestampToMicroseconds(const int64 timestamp);
-  int64 SampleNumberToMicroseconds(const int64 sample_number);
+  int64_t TimestampToMicroseconds(const int64_t timestamp);
+  int64_t SampleNumberToMicroseconds(const int64_t sample_number);
 
   // Returns an error if the sample format in avformat_ctx_.sample_format
   // is not supported.
@@ -161,7 +161,7 @@ class AudioPacketProcessor : public BasePacketProcessor {
   absl::Status ProcessDecodedFrame(const AVPacket& packet) override;
 
   // Corrects PTS for rollover if correction is enabled.
-  int64 MaybeCorrectPtsForRollover(int64 media_pts);
+  int64_t MaybeCorrectPtsForRollover(int64_t media_pts);
 
   // Number of channels to output. This value might be different from
   // the actual number of channels for the current AVPacket, found in
@@ -171,7 +171,7 @@ class AudioPacketProcessor : public BasePacketProcessor {
   // Sample rate of the data to output. This value might be different
   // from the actual sample rate for the current AVPacket, found in
   // avcodec_ctx_->sample_rate.
-  int64 sample_rate_ = -1;
+  int64_t sample_rate_ = -1;
 
   // The time base of audio samples (i.e. the reciprocal of the sample rate).
   AVRational sample_time_base_;
@@ -180,7 +180,7 @@ class AudioPacketProcessor : public BasePacketProcessor {
   Timestamp last_timestamp_;
 
   // The expected sample number based on counting samples.
-  int64 expected_sample_number_ = 0;
+  int64_t expected_sample_number_ = 0;
 
   // Options for the processor.
   AudioStreamOptions options_;

mediapipe/util/image_frame_util.h

@@ -74,14 +74,14 @@ void YUVImageToImageFrameFromFormat(const YUVImage& yuv_image,
 // values used are those from ITU-R BT.601 (which are the same as ITU-R
 // BT.709).  The conversion values are taken from wikipedia and cross
 // checked with other sources.
-void SrgbToMpegYCbCr(const uint8 r, const uint8 g, const uint8 b,  //
-                     uint8* y, uint8* cb, uint8* cr);
+void SrgbToMpegYCbCr(const uint8_t r, const uint8_t g, const uint8_t b,  //
+                     uint8_t* y, uint8_t* cb, uint8_t* cr);
 // Convert MPEG YCbCr values into sRGB values.  See the SrgbToMpegYCbCr()
 // for more notes.  Many MPEG YCbCr values do not correspond directly
 // to an sRGB value.  If the value is invalid it will be clipped to the
 // closest valid value on a per channel basis.
-void MpegYCbCrToSrgb(const uint8 y, const uint8 cb, const uint8 cr,  //
-                     uint8* r, uint8* g, uint8* b);
+void MpegYCbCrToSrgb(const uint8_t y, const uint8_t cb, const uint8_t cr,  //
+                     uint8_t* r, uint8_t* g, uint8_t* b);
 
 // Conversion functions to and from srgb and linear RGB in 16 bits-per-pixel
 // channel.

mediapipe/util/label_map_util.h

@@ -27,7 +27,7 @@ namespace mediapipe {
 // both expected to contain one label per line.
 // Returns an error e.g. if there's a mismatch between the number of labels and
 // display names.
-absl::StatusOr<proto_ns::Map<int64, ::mediapipe::LabelMapItem>>
+absl::StatusOr<proto_ns::Map<int64_t, ::mediapipe::LabelMapItem>>
 BuildLabelMapFromFiles(absl::string_view labels_file_contents,
                        absl::string_view display_names_file);
 

mediapipe/util/time_series_test_util.h

@@ -292,7 +292,7 @@ class TimeSeriesCalculatorTest : public ::testing::Test {
 
   // Overload to allow explicit conversion from int64 to Timestamp
   template <typename T>
-  void AppendInputPacket(const T* payload, const int64 timestamp,
+  void AppendInputPacket(const T* payload, const int64_t timestamp,
                          const size_t input_index = 0) {
     AppendInputPacket(payload, Timestamp(timestamp), input_index);
   }
@@ -305,7 +305,7 @@ class TimeSeriesCalculatorTest : public ::testing::Test {
   }
 
   template <typename T>
-  void AppendInputPacket(const T* payload, const int64 timestamp,
+  void AppendInputPacket(const T* payload, const int64_t timestamp,
                          const std::string& input_tag) {
     AppendInputPacket(payload, Timestamp(timestamp), input_tag);
   }
@@ -450,7 +450,7 @@ class MultiStreamTimeSeriesCalculatorTest
 
   // Overload to allow explicit conversion from int64 to Timestamp
   void AppendInputPacket(const std::vector<Matrix>* input_vector,
-                         const int64 timestamp) {
+                         const int64_t timestamp) {
     AppendInputPacket(input_vector, Timestamp(timestamp));
   }
 

mediapipe/util/time_series_util.h

@@ -39,7 +39,7 @@ namespace time_series_util {
 // function.
 bool LogWarningIfTimestampIsInconsistent(const Timestamp& current_timestamp,
                                          const Timestamp& initial_timestamp,
-                                         int64 cumulative_samples,
+                                         int64_t cumulative_samples,
                                          double sample_rate);
 
 // Returns absl::Status::OK if the header is valid. Otherwise, returns a
@@ -109,11 +109,11 @@ void SetExtensionInHeader(const TimeSeriesHeaderExtensionClass& extension,
 }
 
 // Converts from a time_in_seconds to an integer number of samples.
-int64 SecondsToSamples(double time_in_seconds, double sample_rate);
+int64_t SecondsToSamples(double time_in_seconds, double sample_rate);
 
 // Converts from an integer number of samples to a time duration in seconds
 // spanned by the samples.
-double SamplesToSeconds(int64 num_samples, double sample_rate);
+double SamplesToSeconds(int64_t num_samples, double sample_rate);
 
 }  // namespace time_series_util
 }  // namespace mediapipe

mediapipe/web/graph_runner/graph_runner.ts

@@ -52,6 +52,26 @@ declare global {
  */
 declare function importScripts(...urls: Array<string|URL>): void;
 
+/**
+ * Detects image source size.
+ */
+export function getImageSourceSize(imageSource: ImageSource): [number, number] {
+  if ((imageSource as HTMLVideoElement).videoWidth !== undefined) {
+    const videoElement = imageSource as HTMLVideoElement;
+    return [videoElement.videoWidth, videoElement.videoHeight];
+  } else if ((imageSource as HTMLImageElement).naturalWidth !== undefined) {
+    // TODO: Ensure this works with SVG images
+    const imageElement = imageSource as HTMLImageElement;
+    return [imageElement.naturalWidth, imageElement.naturalHeight];
+  } else if ((imageSource as VideoFrame).displayWidth !== undefined) {
+    const videoFrame = imageSource as VideoFrame;
+    return [videoFrame.displayWidth, videoFrame.displayHeight];
+  } else {
+    const notVideoFrame = imageSource as Exclude<ImageSource, VideoFrame>;
+    return [notVideoFrame.width, notVideoFrame.height];
+  }
+}
+
 /**
  * Simple class to run an arbitrary image-in/image-out MediaPipe graph (i.e.
  * as created by wasm_mediapipe_demo BUILD macro), and either render results
@@ -64,7 +84,7 @@ export class GraphRunner implements GraphRunnerApi {
   //   should be somewhat fixed when we create our .d.ts files.
   readonly wasmModule: WasmModule;
   readonly hasMultiStreamSupport: boolean;
-  autoResizeCanvas: boolean = true;
+  autoResizeCanvas = true;
   audioPtr: number|null;
   audioSize: number;
 
@@ -196,18 +216,7 @@ export class GraphRunner implements GraphRunnerApi {
       gl.pixelStorei(gl.UNPACK_FLIP_Y_WEBGL, false);
     }
 
-    let width, height;
-    if ((imageSource as HTMLVideoElement).videoWidth) {
-      width = (imageSource as HTMLVideoElement).videoWidth;
-      height = (imageSource as HTMLVideoElement).videoHeight;
-    } else if ((imageSource as HTMLImageElement).naturalWidth) {
-      // TODO: Ensure this works with SVG images
-      width = (imageSource as HTMLImageElement).naturalWidth;
-      height = (imageSource as HTMLImageElement).naturalHeight;
-    } else {
-      width = imageSource.width;
-      height = imageSource.height;
-    }
+    const [width, height] = getImageSourceSize(imageSource);
 
     if (this.autoResizeCanvas &&
         (width !== this.wasmModule.canvas.width ||
@@ -295,7 +304,7 @@ export class GraphRunner implements GraphRunnerApi {
    * format).
    *
    * Consumers must deserialize the binary representation themselves as this
-   * avoids addding a direct dependency on the Protobuf JSPB target in the graph
+   * avoids adding a direct dependency on the Protobuf JSPB target in the graph
    * library.
    */
   getCalculatorGraphConfig(

mediapipe/web/graph_runner/graph_runner_api.d.ts

@@ -26,8 +26,8 @@ export {
 /**
  * Valid types of image sources which we can run our GraphRunner over.
  */
-export type ImageSource =
-    HTMLCanvasElement|HTMLVideoElement|HTMLImageElement|ImageData|ImageBitmap;
+export type ImageSource = HTMLCanvasElement|HTMLVideoElement|HTMLImageElement|
+    ImageData|ImageBitmap|VideoFrame;
 
 /**
  * Simple interface for a class to run an arbitrary MediaPipe graph on web, and

package.json

@@ -5,7 +5,7 @@
   "devDependencies": {
     "@bazel/jasmine": "^5.7.2",
     "@bazel/rollup": "^5.7.1",
-    "@bazel/typescript": "^5.7.1",
+    "@bazel/typescript": "^5.8.1",
     "@rollup/plugin-commonjs": "^23.0.2",
     "@rollup/plugin-node-resolve": "^15.0.1",
     "@rollup/plugin-terser": "^0.1.0",
@@ -20,6 +20,6 @@
     "protobufjs-cli": "^1.0.2",
     "rollup": "^2.3.0",
     "ts-protoc-gen": "^0.15.0",
-    "typescript": "^4.8.4"
+    "typescript": "^5.3.3"
   }
 }

version.bzl

@@ -2,4 +2,4 @@
 
 # The next version of MediaPipe (e.g. the version that is currently in development).
 # This version should be bumped after every release.
-MEDIAPIPE_FULL_VERSION = "0.10.9"
+MEDIAPIPE_FULL_VERSION = "0.10.10"