Merge pull request #5028 from kinaryml:python-holistic-landmarker

PiperOrigin-RevId: 594995636

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/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)
+    node_config = calculator_pb2.CalculatorGraphConfig.Node(
-    if not enable_flow_limiting:
-      return calculator_pb2.CalculatorGraphConfig(
-          node=[
-              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=[
+        node=[node_config, flow_limiter],
-            calculator_pb2.CalculatorGraphConfig.Node(
-                calculator=self.task_graph,
-                input_stream=task_subgraph_inputs,
-                output_stream=self.output_streams,
-                options=task_subgraph_options), 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

@@ -58,9 +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",
@@ -142,6 +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",
@@ -194,6 +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",