Extract CPU conversion methods into a separate library & add test

PiperOrigin-RevId: 582966041

mediapipe/calculators/tensor/BUILD

@@ -657,6 +657,7 @@ cc_library(
     }),
     deps = [
         ":tensor_converter_calculator_cc_proto",
+        ":tensor_converter_cpu",
         "//mediapipe/framework:calculator_framework",
         "//mediapipe/framework:port",
         "//mediapipe/framework/formats:image_frame",
@@ -665,6 +666,7 @@ cc_library(
         "//mediapipe/framework/port:ret_check",
         "//mediapipe/framework/port:status",
         "//mediapipe/framework/port:statusor",
+        "//mediapipe/gpu:gpu_buffer",
         "//mediapipe/gpu:gpu_buffer_format",
         "//mediapipe/gpu:gpu_origin_cc_proto",
         "//mediapipe/util:resource_util",
@@ -674,10 +676,17 @@ cc_library(
         "@com_google_absl//absl/log:check",
         "@com_google_absl//absl/status",
         "@com_google_absl//absl/status:statusor",
+        "@com_google_absl//absl/strings",
         "@com_google_absl//absl/strings:str_format",
     ] + select({
         "//mediapipe/gpu:disable_gpu": [],
-        "//conditions:default": ["tensor_converter_calculator_gpu_deps"],
+        "//conditions:default": [
+            "tensor_converter_calculator_gpu_deps",
+            "//mediapipe/gpu:gl_base",
+            "//mediapipe/gpu:gl_calculator_helper",
+            "//mediapipe/gpu:gl_simple_shaders",
+            "//mediapipe/gpu:shader_util",
+        ],
     }) + select({
         "//mediapipe:apple": [
             "//third_party/apple_frameworks:MetalKit",
@@ -687,6 +696,35 @@ cc_library(
     alwayslink = 1,
 )
 
+cc_library(
+    name = "tensor_converter_cpu",
+    srcs = ["tensor_converter_cpu.cc"],
+    hdrs = ["tensor_converter_cpu.h"],
+    deps = [
+        "//mediapipe/framework/formats:image_frame",
+        "//mediapipe/framework/formats:matrix",
+        "//mediapipe/framework/formats:tensor",
+        "//mediapipe/framework/port:ret_check",
+        "//mediapipe/framework/port:status",
+        "@com_google_absl//absl/status",
+        "@com_google_absl//absl/status:statusor",
+    ],
+)
+
+cc_test(
+    name = "tensor_converter_cpu_test",
+    srcs = ["tensor_converter_cpu_test.cc"],
+    deps = [
+        ":tensor_converter_cpu",
+        "//mediapipe/framework/formats:matrix",
+        "//mediapipe/framework/formats:tensor",
+        "//mediapipe/framework/port:gtest",
+        "//mediapipe/framework/port:gtest_main",
+        "//mediapipe/framework/port:status_matchers",
+        "//mediapipe/util:image_test_utils",
+    ],
+)
+
 cc_library(
     name = "tensor_converter_calculator_gpu_deps",
     visibility = ["//visibility:private"],

mediapipe/calculators/tensor/tensor_converter_calculator.cc

@@ -14,6 +14,7 @@
 
 #include <cstdint>
 #include <string>
+#include <utility>
 #include <vector>
 
 #include "absl/log/absl_check.h"
@@ -21,17 +22,21 @@
 #include "absl/status/status.h"
 #include "absl/status/statusor.h"
 #include "absl/strings/str_format.h"
+#include "absl/strings/substitute.h"
 #include "mediapipe/calculators/tensor/tensor_converter_calculator.pb.h"
+#include "mediapipe/calculators/tensor/tensor_converter_cpu.h"
 #include "mediapipe/framework/calculator_framework.h"
 #include "mediapipe/framework/formats/image_frame.h"
 #include "mediapipe/framework/formats/matrix.h"
 #include "mediapipe/framework/formats/tensor.h"
 #include "mediapipe/framework/port.h"
 #include "mediapipe/framework/port/ret_check.h"
+#include "mediapipe/framework/port/status_macros.h"
 #include "mediapipe/gpu/gpu_buffer_format.h"
 #include "mediapipe/gpu/gpu_origin.pb.h"
 
 #if !MEDIAPIPE_DISABLE_GPU
+#include "mediapipe/gpu/gl_base.h"
 #include "mediapipe/gpu/gpu_buffer.h"
 #if MEDIAPIPE_METAL_ENABLED
 #import <CoreVideo/CoreVideo.h>
@@ -94,11 +99,6 @@ absl::StatusOr<bool> ShouldFlipVertically(
   }
 }
 
-typedef Eigen::Matrix<float, Eigen::Dynamic, Eigen::Dynamic, Eigen::RowMajor>
-    RowMajorMatrixXf;
-typedef Eigen::Matrix<float, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor>
-    ColMajorMatrixXf;
-
 constexpr char kImageFrameTag[] = "IMAGE";
 constexpr char kGpuBufferTag[] = "IMAGE_GPU";
 constexpr char kTensorsTag[] = "TENSORS";
@@ -156,10 +156,6 @@ class TensorConverterCalculator : public CalculatorBase {
  private:
   absl::Status InitGpu(CalculatorContext* cc);
   absl::Status LoadOptions(CalculatorContext* cc, bool use_gpu);
-  template <class T>
-  absl::Status NormalizeImage(const ImageFrame& image_frame,
-                              bool flip_vertically, float* tensor_ptr);
-  absl::Status CopyMatrixToTensor(const Matrix& matrix, float* tensor_ptr);
   absl::Status ProcessCPU(CalculatorContext* cc);
   absl::Status ProcessGPU(CalculatorContext* cc);
 
@@ -279,46 +275,19 @@ absl::Status TensorConverterCalculator::ProcessCPU(CalculatorContext* cc) {
     }
     const auto& image_frame =
         cc->Inputs().Tag(kImageFrameTag).Get<ImageFrame>();
-    const int height = image_frame.Height();
+    MP_ASSIGN_OR_RETURN(
-    const int width = image_frame.Width();
+        Tensor output,
-    const int channels = image_frame.NumberOfChannels();
+        ConvertImageFrameToTensorOnCpu(image_frame, *output_range_,
-    const int channels_preserved = std::min(channels, max_num_channels_);
+                                       flip_vertically_, max_num_channels_));
-    const mediapipe::ImageFormat::Format format = image_frame.Format();
+    output_tensors->emplace_back(std::move(output));
-
-    if (!(format == mediapipe::ImageFormat::SRGBA ||
-          format == mediapipe::ImageFormat::SRGB ||
-          format == mediapipe::ImageFormat::GRAY8 ||
-          format == mediapipe::ImageFormat::VEC32F1))
-      RET_CHECK_FAIL() << "Unsupported CPU input format.";
-
-    output_tensors->emplace_back(
-        Tensor::ElementType::kFloat32,
-        Tensor::Shape{1, height, width, channels_preserved});
-    auto cpu_view = output_tensors->back().GetCpuWriteView();
-
-    // Copy image data into tensor.
-    if (image_frame.ByteDepth() == 1) {
-      MP_RETURN_IF_ERROR(NormalizeImage<uint8_t>(image_frame, flip_vertically_,
-                                                 cpu_view.buffer<float>()));
-    } else if (image_frame.ByteDepth() == 4) {
-      MP_RETURN_IF_ERROR(NormalizeImage<float>(image_frame, flip_vertically_,
-                                               cpu_view.buffer<float>()));
-    } else {
-      return absl::InternalError(
-          "Only byte-based (8 bit) and float (32 bit) images supported.");
-    }
   } else if (cc->Inputs().HasTag(kMatrixTag)) {
     if (cc->Inputs().Tag(kMatrixTag).IsEmpty()) {
       return absl::OkStatus();
     }
     const auto& matrix = cc->Inputs().Tag(kMatrixTag).Get<Matrix>();
-    const int height = matrix.rows();
+    MP_ASSIGN_OR_RETURN(Tensor output,
-    const int width = matrix.cols();
+                        ConvertMatrixToTensorOnCpu(matrix, row_major_matrix_));
-    const int channels = 1;
+    output_tensors->emplace_back(std::move(output));
-    output_tensors->emplace_back(Tensor::ElementType::kFloat32,
-                                 Tensor::Shape{1, height, width, channels});
-    MP_RETURN_IF_ERROR(CopyMatrixToTensor(
-        matrix, output_tensors->back().GetCpuWriteView().buffer<float>()));
   } else {
     return absl::OkStatus();
   }
@@ -669,67 +638,4 @@ absl::Status TensorConverterCalculator::LoadOptions(CalculatorContext* cc,
   return absl::OkStatus();
 }
 
-template <class T>
-absl::Status TensorConverterCalculator::NormalizeImage(
-    const ImageFrame& image_frame, bool flip_vertically, float* tensor_ptr) {
-  const int height = image_frame.Height();
-  const int width = image_frame.Width();
-  const int channels = image_frame.NumberOfChannels();
-  const int channels_preserved = std::min(channels, max_num_channels_);
-  const int channels_ignored = channels - channels_preserved;
-
-  if (output_range_.has_value()) {
-    // If the output float range is set and we are not using custom
-    // normalization, normalize the pixel values from [0, 255] to the specified
-    // output range.
-    RET_CHECK_NE(output_range_->first, output_range_->second);
-    const float scale = (output_range_->second - output_range_->first) / 255.0f;
-    const float bias = output_range_->first;
-
-    for (int i = 0; i < height; ++i) {
-      const T* image_ptr = reinterpret_cast<const T*>(
-          image_frame.PixelData() +
-          (flip_vertically ? height - 1 - i : i) * image_frame.WidthStep());
-      for (int j = 0; j < width; ++j) {
-        for (int c = 0; c < channels_preserved; ++c) {
-          *tensor_ptr++ = *image_ptr++ * scale + bias;
-        }
-        image_ptr += channels_ignored;
-      }
-    }
-  } else {
-    // [0,1], scale only (bias == 0)
-    // Verified that there are no precision issues with 1.0f / 255.0f expression
-    const float scale = 1.0f / 255.0f;
-    for (int i = 0; i < height; ++i) {
-      const T* image_ptr = reinterpret_cast<const T*>(
-          image_frame.PixelData() +
-          (flip_vertically ? height - 1 - i : i) * image_frame.WidthStep());
-      for (int j = 0; j < width; ++j) {
-        for (int c = 0; c < channels_preserved; ++c) {
-          *tensor_ptr++ = *image_ptr++ * scale;
-        }
-        image_ptr += channels_ignored;
-      }
-    }
-  }
-
-  return absl::OkStatus();
-}
-
-absl::Status TensorConverterCalculator::CopyMatrixToTensor(const Matrix& matrix,
-                                                           float* tensor_ptr) {
-  if (row_major_matrix_) {
-    auto matrix_map =
-        Eigen::Map<RowMajorMatrixXf>(tensor_ptr, matrix.rows(), matrix.cols());
-    matrix_map = matrix;
-  } else {
-    auto matrix_map =
-        Eigen::Map<ColMajorMatrixXf>(tensor_ptr, matrix.rows(), matrix.cols());
-    matrix_map = matrix;
-  }
-
-  return absl::OkStatus();
-}
-
 }  // namespace mediapipe

mediapipe/calculators/tensor/tensor_converter_cpu.cc

@@ -0,0 +1,145 @@
+// 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/calculators/tensor/tensor_converter_cpu.h"
+
+#include <algorithm>
+#include <cstdint>
+#include <utility>
+
+#include "absl/status/status.h"
+#include "absl/status/statusor.h"
+#include "mediapipe/framework/formats/image_frame.h"
+#include "mediapipe/framework/formats/matrix.h"
+#include "mediapipe/framework/formats/tensor.h"
+#include "mediapipe/framework/port/ret_check.h"
+#include "mediapipe/framework/port/status_macros.h"
+
+namespace mediapipe {
+namespace {
+
+typedef Eigen::Matrix<float, Eigen::Dynamic, Eigen::Dynamic, Eigen::RowMajor>
+    RowMajorMatrixXf;
+typedef Eigen::Matrix<float, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor>
+    ColMajorMatrixXf;
+
+template <class T>
+absl::Status NormalizeImage(const ImageFrame& image_frame, bool flip_vertically,
+                            const std::pair<float, float>& output_range,
+                            int max_num_channels, float* tensor_ptr) {
+  const int height = image_frame.Height();
+  const int width = image_frame.Width();
+  const int channels = image_frame.NumberOfChannels();
+  const int channels_preserved = std::min(channels, max_num_channels);
+  const int channels_ignored = channels - channels_preserved;
+
+  RET_CHECK_NE(output_range.first, output_range.second);
+  const float scale = (output_range.second - output_range.first) / 255.0f;
+  const float bias = output_range.first;
+
+  for (int i = 0; i < height; ++i) {
+    const T* image_ptr = reinterpret_cast<const T*>(
+        image_frame.PixelData() +
+        (flip_vertically ? height - 1 - i : i) * image_frame.WidthStep());
+    for (int j = 0; j < width; ++j) {
+      for (int c = 0; c < channels_preserved; ++c) {
+        *tensor_ptr++ = *image_ptr++ * scale + bias;
+      }
+      image_ptr += channels_ignored;
+    }
+  }
+  return absl::OkStatus();
+}
+
+}  // namespace
+
+absl::Status NormalizeUInt8Image(const ImageFrame& image_frame,
+                                 bool flip_vertically,
+                                 const std::pair<float, float>& output_range,
+                                 int max_num_channels, float* tensor_ptr) {
+  return NormalizeImage<uint8_t>(image_frame, flip_vertically, output_range,
+                                 max_num_channels, tensor_ptr);
+}
+
+absl::Status NormalizeFloatImage(const ImageFrame& image_frame,
+                                 bool flip_vertically,
+                                 const std::pair<float, float>& output_range,
+                                 int max_num_channels, float* tensor_ptr) {
+  return NormalizeImage<float>(image_frame, flip_vertically, output_range,
+                               max_num_channels, tensor_ptr);
+}
+
+absl::Status CopyMatrixToTensor(const Matrix& matrix, bool is_row_major_matrix,
+                                float* tensor_ptr) {
+  if (is_row_major_matrix) {
+    auto matrix_map =
+        Eigen::Map<RowMajorMatrixXf>(tensor_ptr, matrix.rows(), matrix.cols());
+    matrix_map = matrix;
+  } else {
+    auto matrix_map =
+        Eigen::Map<ColMajorMatrixXf>(tensor_ptr, matrix.rows(), matrix.cols());
+    matrix_map = matrix;
+  }
+  return absl::OkStatus();
+}
+
+absl::StatusOr<Tensor> ConvertImageFrameToTensorOnCpu(
+    const ImageFrame& image_frame, const std::pair<float, float>& output_range,
+    bool flip_vertically, int max_num_channels) {
+  const int height = image_frame.Height();
+  const int width = image_frame.Width();
+  const int channels = image_frame.NumberOfChannels();
+  const int channels_preserved = std::min(channels, max_num_channels);
+  const mediapipe::ImageFormat::Format format = image_frame.Format();
+
+  if (!(format == mediapipe::ImageFormat::SRGBA ||
+        format == mediapipe::ImageFormat::SRGB ||
+        format == mediapipe::ImageFormat::GRAY8 ||
+        format == mediapipe::ImageFormat::VEC32F1))
+    RET_CHECK_FAIL() << "Unsupported CPU input format.";
+
+  Tensor output_tensor(Tensor::ElementType::kFloat32,
+                       Tensor::Shape{1, height, width, channels_preserved});
+  auto cpu_view = output_tensor.GetCpuWriteView();
+
+  // Copy image data into tensor.
+  if (image_frame.ByteDepth() == 1) {
+    MP_RETURN_IF_ERROR(NormalizeUInt8Image(image_frame, flip_vertically,
+                                           output_range, max_num_channels,
+                                           cpu_view.buffer<float>()));
+  } else if (image_frame.ByteDepth() == 4) {
+    MP_RETURN_IF_ERROR(NormalizeFloatImage(image_frame, flip_vertically,
+                                           output_range, max_num_channels,
+                                           cpu_view.buffer<float>()));
+  } else {
+    return absl::InternalError(
+        "Only byte-based (8 bit) and float (32 bit) images supported.");
+  }
+  return output_tensor;
+}
+
+absl::StatusOr<Tensor> ConvertMatrixToTensorOnCpu(const Matrix& matrix,
+                                                  bool row_major_matrix) {
+  const int height = matrix.rows();
+  const int width = matrix.cols();
+  const int channels = 1;
+  Tensor output_tensor(Tensor::ElementType::kFloat32,
+                       Tensor::Shape{1, height, width, channels});
+  MP_RETURN_IF_ERROR(
+      CopyMatrixToTensor(matrix, row_major_matrix,
+                         output_tensor.GetCpuWriteView().buffer<float>()));
+  return output_tensor;
+}
+
+}  // namespace mediapipe

mediapipe/calculators/tensor/tensor_converter_cpu.h

@@ -0,0 +1,61 @@
+// 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_CALCULATORS_TENSOR_TENSOR_CONVERTER_CPU_H_
+#define MEDIAPIPE_CALCULATORS_TENSOR_TENSOR_CONVERTER_CPU_H_
+
+#include <utility>
+
+#include "absl/status/status.h"
+#include "absl/status/statusor.h"
+#include "mediapipe/framework/formats/image_frame.h"
+#include "mediapipe/framework/formats/matrix.h"
+#include "mediapipe/framework/formats/tensor.h"
+
+namespace mediapipe {
+
+// Converts an ImageFrame to a vector of Tensors.
+// @flip_vertically enables to flip the image during conversion.
+// @max_num_channels can be used to reserve extra channels in the output
+// tensors.
+// Returns output Tensor.
+absl::StatusOr<Tensor> ConvertImageFrameToTensorOnCpu(
+    const ImageFrame& image_frame, const std::pair<float, float>& output_range,
+    bool flip_vertically, int max_num_channels);
+
+// Converts a Matrix to a vector of Tensors.
+// @row_major_matrix defines the ordering in the input matrix.
+// @max_num_channels can be used to reserve extra channels in the output
+// tensors.
+// Returns output Tensor.
+absl::StatusOr<Tensor> ConvertMatrixToTensorOnCpu(const Matrix& matrix,
+                                                  bool row_major_matrix);
+
+// For testing only below.
+absl::Status NormalizeUInt8Image(const ImageFrame& image_frame,
+                                 bool flip_vertically,
+                                 const std::pair<float, float>& output_range,
+                                 int max_num_channels, float* tensor_ptr);
+
+absl::Status NormalizeFloatImage(const ImageFrame& image_frame,
+                                 bool flip_vertically,
+                                 const std::pair<float, float>& output_range,
+                                 int max_num_channels, float* tensor_ptr);
+
+absl::Status CopyMatrixToTensor(const Matrix& matrix, bool is_row_major_matrix,
+                                float* tensor_ptr);
+
+}  // namespace mediapipe
+
+#endif  // MEDIAPIPE_CALCULATORS_TENSOR_TENSOR_CONVERTER_CPU_H_

mediapipe/calculators/tensor/tensor_converter_cpu_test.cc

@@ -0,0 +1,175 @@
+// 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/calculators/tensor/tensor_converter_cpu.h"
+
+#include <cstdint>
+#include <utility>
+#include <vector>
+
+#include "mediapipe/framework/formats/matrix.h"
+#include "mediapipe/framework/formats/tensor.h"
+#include "mediapipe/framework/port/gmock.h"
+#include "mediapipe/framework/port/gtest.h"
+#include "mediapipe/framework/port/status_matchers.h"
+#include "mediapipe/util/image_test_utils.h"
+
+namespace mediapipe {
+namespace {
+
+Matrix CreateTestMatrix(int num_rows, int num_columns) {
+  Matrix matrix(num_rows, num_columns);
+  for (int r = 0; r < num_rows; ++r) {
+    for (int c = 0; c < num_columns; ++c) {
+      matrix(r, c) = r * num_columns + c;
+    }
+  }
+  return matrix;
+}
+
+TEST(TensorConverterCpuTest, ShouldCopyMatrixInRowMajorFormatToTensor) {
+  auto test_matrix = CreateTestMatrix(/* num_rows=*/3, /*num_columns=*/4);
+  std::vector<float> tensor_data(test_matrix.size(), 0.0f);
+
+  MP_EXPECT_OK(CopyMatrixToTensor(test_matrix, /*is_row_major_matrix=*/true,
+                                  tensor_data.data()));
+
+  for (int i = 0; i < tensor_data.size(); ++i) {
+    const int row = i / test_matrix.cols();
+    const int column = i % test_matrix.cols();
+    EXPECT_FLOAT_EQ(tensor_data[i], (test_matrix)(row, column));
+  }
+}
+
+TEST(TensorConverterCpuTest, ShouldCopyMatrixInColumnMajorFormatToTensor) {
+  auto test_matrix = CreateTestMatrix(/*num_rows=*/3, /*num_columns=*/4);
+  std::vector<float> tensor_data(test_matrix.size(), 0.0f);
+
+  MP_EXPECT_OK(CopyMatrixToTensor(test_matrix, /*is_row_major_matrix=*/false,
+                                  tensor_data.data()));
+
+  for (int i = 0; i < tensor_data.size(); ++i) {
+    const int row = i % test_matrix.rows();
+    const int column = i / test_matrix.rows();
+    EXPECT_FLOAT_EQ(tensor_data[i], (test_matrix)(row, column));
+  }
+}
+
+TEST(TensorConverterCpuTest, ShouldNormalizeGrey8ImageWithDefaultRange) {
+  auto grey8_image_frame = CreateTestGrey8ImageFrame(/*width=*/3, /*height=*/4);
+  std::vector<float> tensor_data(
+      grey8_image_frame.Width() * grey8_image_frame.Height(), 0.0f);
+
+  MP_EXPECT_OK(NormalizeUInt8Image(grey8_image_frame, /*flip_vertically=*/false,
+                                   {0.0f, 1.0f}, /*num_tensor_channels=*/1,
+                                   tensor_data.data()));
+
+  for (int i = 0; i < tensor_data.size(); ++i) {
+    EXPECT_FLOAT_EQ(
+        tensor_data[i],
+        static_cast<uint8_t>(grey8_image_frame.PixelData()[i]) / 255.0f);
+  }
+}
+
+TEST(TensorConverterCpuTest, ShouldNormalizeGrey8ImageWithSpecifiedRange) {
+  auto grey8_image_frame = CreateTestGrey8ImageFrame(/*width=*/3, /*height=*/4);
+  std::vector<float> tensor_data(
+      grey8_image_frame.Width() * grey8_image_frame.Height(), 0.0f);
+  const auto range = std::make_pair(2.0f, 3.0f);
+
+  MP_EXPECT_OK(
+      NormalizeUInt8Image(grey8_image_frame, /*flip_vertically=*/false, range,
+                          /*num_tensor_channels=*/1, tensor_data.data()));
+
+  for (int i = 0; i < tensor_data.size(); ++i) {
+    EXPECT_FLOAT_EQ(tensor_data[i],
+                    static_cast<uint8_t>(grey8_image_frame.PixelData()[i]) /
+                            255.0f * (range.second - range.first) +
+                        range.first);
+  }
+}
+
+TEST(TensorConverterCpuTest, ShouldNormalizeGrey8ImageFlipped) {
+  auto grey8_image_frame = CreateTestGrey8ImageFrame(/*width=*/3, /*height=*/4);
+  std::vector<float> tensor_data(
+      grey8_image_frame.Width() * grey8_image_frame.Height(), 0.0f);
+
+  MP_EXPECT_OK(NormalizeUInt8Image(grey8_image_frame, /*flip_vertically=*/true,
+                                   {0.0f, 1.0f}, /*num_tensor_channels=*/1,
+                                   tensor_data.data()));
+
+  for (int i = 0; i < tensor_data.size(); ++i) {
+    const int x = i % grey8_image_frame.Width();
+    const int y = i / grey8_image_frame.Width();
+    const int flipped_y = grey8_image_frame.Height() - y - 1;
+
+    const int index = flipped_y * grey8_image_frame.Width() + x;
+    EXPECT_FLOAT_EQ(
+        tensor_data[index],
+        static_cast<uint8_t>(grey8_image_frame.PixelData()[i]) / 255.0f);
+  }
+}
+
+TEST(TensorConverterCpuTest, ShouldNormalizeFloatImageWithDefaultRange) {
+  auto float_image_frame =
+      CreateTestFloat32ImageFrame(/*width=*/3, /*height=*/4);
+  std::vector<float> tensor_data(
+      float_image_frame.Width() * float_image_frame.Height(), 0.0f);
+
+  MP_EXPECT_OK(NormalizeFloatImage(float_image_frame, /*flip_vertically=*/false,
+                                   {0.0f, 1.0f}, /*num_tensor_channels=*/1,
+                                   tensor_data.data()));
+
+  for (int i = 0; i < tensor_data.size(); ++i) {
+    EXPECT_FLOAT_EQ(tensor_data[i], reinterpret_cast<const float*>(
+                                        float_image_frame.PixelData())[i] /
+                                        255.0f);
+  }
+}
+
+TEST(TensorConverterCpuTest, ConvertImageFrameToTensorOnCpu) {
+  auto grey8_image_frame = CreateTestGrey8ImageFrame(/*width=*/3, /*height=*/4);
+
+  MP_ASSERT_OK_AND_ASSIGN(Tensor output, ConvertImageFrameToTensorOnCpu(
+                                             grey8_image_frame, {0.0f, 1.0f},
+                                             /*flip_vertically=*/false,
+                                             /*max_num_channels=*/1));
+
+  const auto cpu_read_view = output.GetCpuReadView();
+  const float* tensor_ptr = cpu_read_view.buffer<float>();
+  for (int i = 0; i < grey8_image_frame.Width() * grey8_image_frame.Height();
+       ++i) {
+    EXPECT_FLOAT_EQ(
+        tensor_ptr[i],
+        static_cast<uint8_t>(grey8_image_frame.PixelData()[i]) / 255.0);
+  }
+}
+
+TEST(TensorConverterCpuTest, ConvertMatrixToTensorOnCpu) {
+  auto test_matrix = CreateTestMatrix(/*num_rows=*/3, /*num_columns=*/4);
+
+  MP_ASSERT_OK_AND_ASSIGN(
+      Tensor output, ConvertMatrixToTensorOnCpu(test_matrix,
+                                                /*row_major_matrix=*/false));
+
+  const auto cpu_read_view = output.GetCpuReadView();
+  const float* tensor_ptr = cpu_read_view.buffer<float>();
+  for (int i = 0; i < test_matrix.size(); ++i) {
+    EXPECT_FLOAT_EQ(tensor_ptr[i], test_matrix.data()[i]);
+  }
+}
+
+}  // namespace
+
+}  // namespace mediapipe

mediapipe/util/image_test_utils.cc

@@ -17,6 +17,34 @@
 
 namespace mediapipe {
 
+namespace {
+
+template <ImageFormat::Format Format, typename DataType>
+ImageFrame CreateTestImageFrame(int width, int height, DataType max_value) {
+  ImageFrame image_frame(Format, width, height,
+                         /*alignment_boundary=*/1);
+  const int num_channels = image_frame.NumberOfChannels();
+  const float num_values = width * height * num_channels;
+  uint8_t* const data_ptr =
+      reinterpret_cast<uint8_t*>(image_frame.MutablePixelData());
+  for (int y = 0; y < height; ++y) {
+    uint8_t* const row = data_ptr + image_frame.WidthStep() * y;
+    for (int x = 0; x < width; ++x) {
+      DataType* pixel = reinterpret_cast<DataType*>(row) + x * num_channels;
+      for (int c = 0; c < num_channels; ++c) {
+        // Fill pixel channel with a value in [0:max_value] range.
+        pixel[c] =
+            static_cast<DataType>(static_cast<float>(y * width * num_channels +
+                                                     x * num_channels + c) /
+                                  num_values * max_value);
+      }
+    }
+  }
+  return image_frame;
+}
+
+}  // namespace
+
 cv::Mat GetRgb(const std::string& path) {
   cv::Mat bgr = cv::imread(path);
   cv::Mat rgb;
@@ -71,4 +99,14 @@ cv::Mat RgbaToBgr(cv::Mat rgba) {
   return bgra;
 }
 
+ImageFrame CreateTestFloat32ImageFrame(int width, int height) {
+  return CreateTestImageFrame<ImageFormat::VEC32F1, float>(width, height,
+                                                           /*max_value=*/1.0f);
+}
+
+ImageFrame CreateTestGrey8ImageFrame(int width, int height) {
+  return CreateTestImageFrame<ImageFormat::GRAY8, uint8_t>(width, height,
+                                                           /*max_value=*/255);
+}
+
 }  // namespace mediapipe

mediapipe/util/image_test_utils.h

@@ -4,6 +4,7 @@
 #include <string>
 
 #include "mediapipe/framework/formats/image_format.pb.h"
+#include "mediapipe/framework/formats/image_frame.h"
 #include "mediapipe/framework/packet.h"
 #include "mediapipe/framework/port/opencv_core_inc.h"
 
@@ -30,6 +31,12 @@ Packet MakeImagePacket(cv::Mat input, int timestamp = 0);
 // Converts RGBA Mat to BGR.
 cv::Mat RgbaToBgr(cv::Mat rgba);
 
+// Generates single-channel float32 ImageFrame with increasing [0,1] values.
+ImageFrame CreateTestFloat32ImageFrame(int width, int height);
+
+// Generates single-channel uint8 ImageFrame with increasing [0,255] values.
+ImageFrame CreateTestGrey8ImageFrame(int width, int height);
+
 }  // namespace mediapipe
 
 #endif  // MEDIAPIPE_UTIL_IMAGE_TEST_UTILS_H_