// 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.
#include <cstdint>
#include <memory>
#include <utility>
#include <vector>
#include "absl/log/absl_check.h"
#include "absl/log/absl_log.h"
#include "absl/status/status.h"
#include "absl/status/statusor.h"
#include "absl/strings/str_format.h"
#include "absl/types/optional.h"
#include "mediapipe/calculators/tensor/tensor_converter_calculator.pb.h"
#include "mediapipe/calculators/tensor/tensor_converter_cpu.h"
#include "mediapipe/calculators/tensor/tensor_converter_gpu.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/memory_manager.h"
#include "mediapipe/framework/memory_manager_service.h"
#include "mediapipe/framework/port.h"
#include "mediapipe/framework/port/ret_check.h"
#include "mediapipe/framework/port/status_macros.h"
#include "mediapipe/gpu/gpu_origin.pb.h"
#if !MEDIAPIPE_DISABLE_GPU
#include "mediapipe/gpu/gpu_buffer.h"
#include "mediapipe/gpu/gpu_buffer_format.h"
#include "mediapipe/gpu/gpu_origin.pb.h"
#if MEDIAPIPE_METAL_ENABLED
#include "mediapipe/calculators/tensor/tensor_converter_metal.h"
#import "mediapipe/gpu/MPPMetalHelper.h"
#elif MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
#include "mediapipe/gpu/gl_calculator_helper.h"
#if MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_31
#include "mediapipe/calculators/tensor/tensor_converter_gl31.h"
#else
#include "mediapipe/calculators/tensor/tensor_converter_gl30.h"
#endif // MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_31
#endif // MEDIAPIPE_METAL_ENABLED
#endif // !MEDIAPIPE_DISABLE_GPU
namespace {
// Commonly used to compute the number of blocks to launch in a kernel.
int NumGroups(const int size, const int group_size) { // NOLINT
return (size + group_size - 1) / group_size;
}
absl::StatusOr<bool> ShouldFlipVertically(
const mediapipe::TensorConverterCalculatorOptions& options, bool use_gpu) {
if (options.has_flip_vertically() && options.has_gpu_origin()) {
return absl::FailedPreconditionError(absl::StrFormat(
"Cannot specify both flip_vertically and gpu_origin options"));
}
if (!options.has_gpu_origin()) {
// Fall back to flip_vertically.
return options.flip_vertically();
}
// Warn if gpu_origin is specified with a CPU input image.
// Those are always TOP_LEFT, so no flipping is necessary.
if (!use_gpu) {
ABSL_LOG(WARNING)
<< "Ignoring gpu_origin option since IMAGE_GPU input is not specified";
return false;
}
switch (options.gpu_origin()) {
case mediapipe::GpuOrigin::TOP_LEFT:
return false;
case mediapipe::GpuOrigin::DEFAULT:
case mediapipe::GpuOrigin::CONVENTIONAL:
// TOP_LEFT on Metal, BOTTOM_LEFT on OpenGL.
#ifdef __APPLE__
return false;
#else
return true;
#endif
default:
return absl::InvalidArgumentError(
absl::StrFormat("Unhandled GPU origin %i", options.gpu_origin()));
}
}
constexpr char kImageFrameTag[] = "IMAGE";
constexpr char kGpuBufferTag[] = "IMAGE_GPU";
constexpr char kTensorsTag[] = "TENSORS";
constexpr char kTensorTag[] = "TENSOR";
constexpr char kMatrixTag[] = "MATRIX";
constexpr std::pair<float, float> kDefaultOutputRange = {0.0f, 1.0f};
} // namespace
namespace mediapipe {
// Calculator for normalizing and converting an ImageFrame, GpuBuffer or Matrix
// into a Tensor.
//
// This calculator is designed to be used with the TfLiteInferenceCalculator,
// as a pre-processing step for calculator inputs.
//
// IMAGE and IMAGE_GPU inputs are normalized to [-1,1] (default) or [0,1],
// specified by options (unless outputting a quantized tensor).
//
// Input:
// One of the following tags:
// IMAGE - ImageFrame (assumed to be 8-bit or 32-bit data).
// IMAGE_GPU - GpuBuffer (assumed to be RGBA or RGB GL texture).
// MATRIX - Matrix.
//
// Output:
// One of the following tags:
// TENSORS - Vector of Tensors of type kFloat32. The resource type used:
// - MTLBuffer if Metal API is available
// - SSBO if Metal is unavailable and OpenGL ES 3.1 is available
// - Texture2D if Metal and GLES 3.1 are not available and GLES 3.0 is.
// TENSOR - Tensor of type kFloat32. Resource type same as in TENSORS
//
// Example use:
// node {
// calculator: "TensorConverterCalculator"
// input_stream: "IMAGE:input_image"
// output_stream: "TENSORS:image_tensor"
// options: {
// [mediapipe.TensorConverterCalculatorOptions.ext] {
// zero_center: true
// }
// }
// }
//
// IMPORTANT Notes:
// GPU tensors are currently only supported on mobile platforms.
class TensorConverterCalculator : public CalculatorBase {
public:
static absl::Status GetContract(CalculatorContract* cc);
absl::Status Open(CalculatorContext* cc) override;
absl::Status Process(CalculatorContext* cc) override;
absl::Status Close(CalculatorContext* cc) override;
private:
absl::Status InitGpu(CalculatorContext* cc);
absl::Status LoadOptions(CalculatorContext* cc, bool use_gpu);
absl::StatusOr<std::optional<Tensor>> ProcessCPU(CalculatorContext* cc);
absl::StatusOr<std::optional<Tensor>> ProcessGPU(CalculatorContext* cc);
#if MEDIAPIPE_METAL_ENABLED
MPPMetalHelper* gpu_helper_ = nullptr;
#elif MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
GlCalculatorHelper gpu_helper_;
#endif // MEDIAPIPE_METAL_ENABLED
bool initialized_ = false;
bool use_gpu_ = false;
std::optional<std::pair<float, float>> output_range_;
bool flip_vertically_ = false;
bool row_major_matrix_ = false;
int max_num_channels_ = 3;
std::unique_ptr<TensorConverterGpu> tensor_converter_gpu_;
// Enable pooling of AHWBs in Tensor instances.
MemoryManager* memory_manager_ = nullptr;
};
REGISTER_CALCULATOR(TensorConverterCalculator);
absl::Status TensorConverterCalculator::GetContract(CalculatorContract* cc) {
// Confirm only one of the input streams is present.
RET_CHECK(static_cast<int>(cc->Inputs().HasTag(kImageFrameTag)) +
static_cast<int>(cc->Inputs().HasTag(kGpuBufferTag)) +
static_cast<int>(cc->Inputs().HasTag(kMatrixTag)) ==
1)
<< "Only one input tag of {IMAGE, IMAGE_GPU, MATRIX} may be specified";
if (cc->Inputs().HasTag(kImageFrameTag)) {
cc->Inputs().Tag(kImageFrameTag).Set<ImageFrame>();
}
if (cc->Inputs().HasTag(kMatrixTag)) {
cc->Inputs().Tag(kMatrixTag).Set<Matrix>();
}
cc->UseService(kMemoryManagerService).Optional();
#if !MEDIAPIPE_DISABLE_GPU
if (cc->Inputs().HasTag(kGpuBufferTag)) {
cc->Inputs().Tag(kGpuBufferTag).Set<mediapipe::GpuBuffer>();
#if MEDIAPIPE_METAL_ENABLED
MP_RETURN_IF_ERROR([MPPMetalHelper updateContract:cc]);
#elif MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
MP_RETURN_IF_ERROR(GlCalculatorHelper::UpdateContract(cc));
#endif // MEDIAPIPE_METAL_ENABLED
}
#endif // !MEDIAPIPE_DISABLE_GPU
RET_CHECK(cc->Outputs().HasTag(kTensorsTag) ^
cc->Outputs().HasTag(kTensorTag))
<< "One and only one of TENSOR or TENSORS should be set";
if (cc->Outputs().HasTag(kTensorsTag)) {
cc->Outputs().Tag(kTensorsTag).Set<std::vector<Tensor>>();
}
if (cc->Outputs().HasTag(kTensorTag)) {
cc->Outputs().Tag(kTensorTag).Set<Tensor>();
}
return absl::OkStatus();
}
absl::Status TensorConverterCalculator::Open(CalculatorContext* cc) {
if (cc->Service(kMemoryManagerService).IsAvailable()) {
memory_manager_ = &cc->Service(kMemoryManagerService).GetObject();
}
cc->SetOffset(TimestampDiff(0));
#if !MEDIAPIPE_DISABLE_GPU
if (cc->Inputs().HasTag(kGpuBufferTag)) {
use_gpu_ = true;
#if MEDIAPIPE_METAL_ENABLED
gpu_helper_ = [[MPPMetalHelper alloc] initWithCalculatorContext:cc];
RET_CHECK(gpu_helper_);
#elif MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
MP_RETURN_IF_ERROR(gpu_helper_.Open(cc));
#endif // MEDIAPIPE_METAL_ENABLED
}
#endif // !MEDIAPIPE_DISABLE_GPU
MP_RETURN_IF_ERROR(LoadOptions(cc, use_gpu_));
return absl::OkStatus();
}
absl::Status TensorConverterCalculator::Process(CalculatorContext* cc) {
MP_ASSIGN_OR_RETURN(auto maybe_tensor,
[&]() -> absl::StatusOr<std::optional<Tensor>> {
if (use_gpu_) {
if (cc->Inputs().Tag(kGpuBufferTag).IsEmpty()) {
return std::nullopt;
}
// Convert to GPU tensors type.
return ProcessGPU(cc);
} else {
// Convert to CPU tensors or Matrix type.
return ProcessCPU(cc);
}
}());
if (maybe_tensor) {
if (cc->Outputs().HasTag(kTensorsTag)) {
auto output = std::make_unique<std::vector<Tensor>>();
output->push_back(*std::move(maybe_tensor));
cc->Outputs()
.Tag(kTensorsTag)
.Add(output.release(), cc->InputTimestamp());
} else {
auto output = std::make_unique<Tensor>(*std::move(maybe_tensor));
cc->Outputs().Tag(kTensorTag).Add(output.release(), cc->InputTimestamp());
}
}
return absl::OkStatus();
}
absl::Status TensorConverterCalculator::Close(CalculatorContext* cc) {
#if !MEDIAPIPE_DISABLE_GPU
if (use_gpu_) {
#if MEDIAPIPE_METAL_ENABLED
tensor_converter_gpu_.reset();
#elif MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
gpu_helper_.RunInGlContext([this] { tensor_converter_gpu_.reset(); });
#endif // MEDIAPIPE_METAL_ENABLED
}
#endif // !MEDIAPIPE_DISABLE_GPU
return absl::OkStatus();
}
absl::StatusOr<std::optional<Tensor>> TensorConverterCalculator::ProcessCPU(
CalculatorContext* cc) {
if (cc->Inputs().HasTag(kImageFrameTag)) {
if (cc->Inputs().Tag(kImageFrameTag).IsEmpty()) {
return std::nullopt;
}
const auto& image_frame =
cc->Inputs().Tag(kImageFrameTag).Get<ImageFrame>();
MP_ASSIGN_OR_RETURN(
Tensor output,
ConvertImageFrameToTensorOnCpu(
image_frame,
output_range_.has_value() ? output_range_.value()
: kDefaultOutputRange,
flip_vertically_, max_num_channels_, memory_manager_));
return std::move(output);
} else if (cc->Inputs().HasTag(kMatrixTag)) {
if (cc->Inputs().Tag(kMatrixTag).IsEmpty()) {
return std::nullopt;
}
const auto& matrix = cc->Inputs().Tag(kMatrixTag).Get<Matrix>();
MP_ASSIGN_OR_RETURN(
Tensor output,
ConvertMatrixToTensorOnCpu(matrix, row_major_matrix_, memory_manager_));
return std::move(output);
} else {
return std::nullopt;
}
}
absl::StatusOr<std::optional<Tensor>> TensorConverterCalculator::ProcessGPU(
CalculatorContext* cc) {
#if !MEDIAPIPE_DISABLE_GPU
if (!initialized_) {
MP_RETURN_IF_ERROR(InitGpu(cc));
initialized_ = true;
}
const auto& input =
cc->Inputs().Tag(kGpuBufferTag).Get<mediapipe::GpuBuffer>();
#if MEDIAPIPE_METAL_ENABLED
Tensor output = tensor_converter_gpu_->Convert(input);
return std::move(output);
#elif MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
std::optional<Tensor> output;
MP_RETURN_IF_ERROR(
gpu_helper_.RunInGlContext([this, &output, &input]() -> absl::Status {
output = tensor_converter_gpu_->Convert(input);
return absl::OkStatus();
}));
return std::move(output);
#endif // MEDIAPIPE_METAL_ENABLED
#else
RET_CHECK_FAIL() << "GPU processing is not enabled.";
#endif // !MEDIAPIPE_DISABLE_GPU
return std::nullopt;
}
absl::Status TensorConverterCalculator::InitGpu(CalculatorContext* cc) {
#if !MEDIAPIPE_DISABLE_GPU
// Get input image sizes.
const auto& input =
cc->Inputs().Tag(kGpuBufferTag).Get<mediapipe::GpuBuffer>();
mediapipe::GpuBufferFormat format = input.format();
const bool include_alpha = (max_num_channels_ == 4);
const bool single_channel = (max_num_channels_ == 1);
RET_CHECK(format == mediapipe::GpuBufferFormat::kBGRA32 ||
format == mediapipe::GpuBufferFormat::kRGB24 ||
format == mediapipe::GpuBufferFormat::kRGBA32 ||
format == mediapipe::GpuBufferFormat::kRGBAFloat128 ||
format == mediapipe::GpuBufferFormat::kRGBAHalf64 ||
format == mediapipe::GpuBufferFormat::kGrayFloat32 ||
format == mediapipe::GpuBufferFormat::kGrayHalf16 ||
format == mediapipe::GpuBufferFormat::kOneComponent8)
<< "Unsupported GPU input format: " << static_cast<uint32_t>(format);
if (include_alpha) {
RET_CHECK(format == mediapipe::GpuBufferFormat::kBGRA32 ||
format == mediapipe::GpuBufferFormat::kRGBA32 ||
format == mediapipe::GpuBufferFormat::kRGBAFloat128 ||
format == mediapipe::GpuBufferFormat::kRGBAHalf64)
<< "Num input channels is less than desired output, input format: "
<< static_cast<uint32_t>(format);
}
#if MEDIAPIPE_METAL_ENABLED
MP_ASSIGN_OR_RETURN(
tensor_converter_gpu_,
CreateTensorConverterMetal(gpu_helper_, memory_manager_, output_range_,
include_alpha, single_channel,
flip_vertically_, max_num_channels_));
#elif MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
MP_RETURN_IF_ERROR(gpu_helper_.RunInGlContext(
[this, &input, &include_alpha, &single_channel]() -> absl::Status {
#if MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_31
MP_ASSIGN_OR_RETURN(
tensor_converter_gpu_,
CreateTensorConverterGl31(
gpu_helper_, memory_manager_, input.width(), input.height(),
output_range_, include_alpha, single_channel, flip_vertically_,
max_num_channels_));
#else
MP_ASSIGN_OR_RETURN(
tensor_converter_gpu_,
CreateTensorConverterGl30(
gpu_helper_, memory_manager_, input.width(), input.height(),
output_range_, include_alpha, single_channel, flip_vertically_,
max_num_channels_));
#endif // MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_31
return absl::OkStatus();
}));
#endif // MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_30
#endif // !MEDIAPIPE_DISABLE_GPU
return absl::OkStatus();
}
absl::Status TensorConverterCalculator::LoadOptions(CalculatorContext* cc,
bool use_gpu) {
// Get calculator options specified in the graph.
const auto& options =
cc->Options<::mediapipe::TensorConverterCalculatorOptions>();
// if zero_center, set output float range to match [-1, 1] as specified in
// calculator proto.
if (options.zero_center()) {
output_range_.emplace(std::pair<float, float>(-1.0, 1.0));
}
// Custom output_tensor_float_range values.
// If the float range is specified in pb text, use the specified values
// instead.
if (options.has_output_tensor_float_range()) {
output_range_.emplace(options.output_tensor_float_range().min(),
options.output_tensor_float_range().max());
ABSL_CHECK_GT(output_range_->second, output_range_->first);
}
// Custom div and sub values.
if (options.use_custom_normalization()) {
output_range_.emplace(std::pair<float, float>(
-options.custom_sub(),
-options.custom_sub() + 255.0 / options.custom_div()));
}
// Get y-flip mode.
MP_ASSIGN_OR_RETURN(flip_vertically_, ShouldFlipVertically(options, use_gpu));
// Get row_major_matrix mode.
row_major_matrix_ = options.row_major_matrix();
// Get desired way to handle input channels.
max_num_channels_ = options.max_num_channels();
ABSL_CHECK_GE(max_num_channels_, 1);
ABSL_CHECK_LE(max_num_channels_, 4);
ABSL_CHECK_NE(max_num_channels_, 2);
return absl::OkStatus();
}
} // namespace mediapipe
Codebase Browser
chromium