chromium/components/segmentation_platform/internal/execution/optimization_guide/segmentation_model_executor.h

// Copyright 2021 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#ifndef COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_EXECUTION_OPTIMIZATION_GUIDE_SEGMENTATION_MODEL_EXECUTOR_H_
#define COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_EXECUTION_OPTIMIZATION_GUIDE_SEGMENTATION_MODEL_EXECUTOR_H_

#include <memory>
#include <vector>

#include "components/optimization_guide/core/base_model_executor.h"
#include "components/segmentation_platform/public/model_provider.h"
#include "components/segmentation_platform/public/proto/segmentation_platform.pb.h"

struct TfLiteTensor;

namespace segmentation_platform {

// Provides a framework for executing a particular segmentation model.
// It requires the loaded TensorFlow Lite model to use a single one dimensional
// tensor of floats as the input and a single output tensor with a single float.
// The length of the vector of floats given as input to execution needs to
// exactly match the length of the one-dimensional input tensor.
// Since the shape of the inputs and outputs across all segmentation models are
// the same, this class can be re-used across all the segmentation model
// executors.
class SegmentationModelExecutor : public optimization_guide::BaseModelExecutor<
                                      ModelProvider::Response,
                                      const ModelProvider::Request&> { … };

}  // namespace segmentation_platform

#endif  // COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_EXECUTION_OPTIMIZATION_GUIDE_SEGMENTATION_MODEL_EXECUTOR_H_