树莓派上的tensorflow lite量化ssd对象检测

时间:2018-10-30 21:38:51

标签: c++ raspberry-pi tensorflow-lite

在使用C ++的带有tensorflow lite的树莓派上,对象检测无法按预期工作。我的代码可以编译并运行,但是输出似乎从未正确填充。我会丢失任何依赖项还是访问错误的结果?

我遵循了以下教程: https://medium.com/tensorflow/training-and-serving-a-realtime-mobile-object-detector-in-30-minutes-with-cloud-tpus-b78971cf1193

并从以下位置获得detect.tflite模型: https://storage.googleapis.com/download.tensorflow.org/models/tflite/pets_ssd_mobilenet_v1_0.75_quant_2018_06_29.zip

我已经为树莓派编译了tensorflow lite和opencv,并修改了minimal.cc以读取图像并进行如下推断:

/* Copyright 2018 The TensorFlow Authors. All Rights Reserved.

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 <cstdio>
#include "tensorflow/contrib/lite/interpreter.h"
#include "tensorflow/contrib/lite/kernels/register.h"
#include "tensorflow/contrib/lite/model.h"
#include "tensorflow/contrib/lite/optional_debug_tools.h"

#include <iostream>
#include "opencv2/core.hpp"
#include "opencv2/imgcodecs.hpp"

// This is an example that is minimal to read a model
// from disk and perform inference.
//
// Usage: detect <tflite model> <image filepath>

using namespace tflite;

#define TFLITE_MINIMAL_CHECK(x)                              \
  if (!(x)) {                                                \
    fprintf(stderr, "Error at %s:%d\n", __FILE__, __LINE__); \
    exit(1);                                                 \
  }

int main(int argc, char* argv[]) {
  if(argc != 3) {
    fprintf(stderr, "minimal <tflite model filepath> <image filepath>\n");
    return 1;
  } else {
      fprintf(stdout, "Reading model from %s\n", argv[1]);
      fprintf(stdout, "Reading image from %s\n", argv[2]);
  }
  const char* model_filename = argv[1];
  const char* image_filename = argv[2];

  // Load model
  std::unique_ptr<tflite::FlatBufferModel> model =
      tflite::FlatBufferModel::BuildFromFile(model_filename);
  TFLITE_MINIMAL_CHECK(model != nullptr);

  // Build the interpreter
  tflite::ops::builtin::BuiltinOpResolver resolver;
  InterpreterBuilder builder(*model.get(), resolver);
  std::unique_ptr<Interpreter> interpreter;
  builder(&interpreter);
  TFLITE_MINIMAL_CHECK(interpreter != nullptr);

  // Allocate tensor buffers.
  TFLITE_MINIMAL_CHECK(interpreter->AllocateTensors() == kTfLiteOk);
  printf("=== Pre-invoke Interpreter State ===\n");
  tflite::PrintInterpreterState(interpreter.get());

  // Fill input buffers
  // TODO(user): Insert code to fill input tensors
  cv::Mat img = cv::imread(image_filename);
  //std::cout << "before: " << interpreter->typed_input_tensor<uchar>(0) << std::endl;
  //std::cout << "image: " << img.data << std::endl;
  memcpy(interpreter->typed_input_tensor<uchar>(0), img.data, img.total() * img.elemSize());
  //std::cout << "after: " << interpreter->typed_input_tensor<uchar>(0) << std::endl;

  // Run inference
  TFLITE_MINIMAL_CHECK(interpreter->Invoke() == kTfLiteOk);
  printf("\n\n=== Post-invoke Interpreter State ===\n");
  tflite::PrintInterpreterState(interpreter.get());

  // Read output buffers
  // TODO(user): Insert getting data out code.
  cv::Mat results0(10, 4, CV_8U);
  cv::Mat results1(1, 10, CV_8U);
  cv::Mat results2(1, 10, CV_8U);
  cv::Mat results3(1, 1, CV_8U);

  results0.data = interpreter->typed_output_tensor<uchar>(0);
  results1.data = interpreter->typed_output_tensor<uchar>(1);
  results2.data = interpreter->typed_output_tensor<uchar>(2);
  results3.data = interpreter->typed_output_tensor<uchar>(3);

  std::cout << "results 0: " << results0 << std::endl;
  std::cout << "results 1: " << results1 << std::endl;
  std::cout << "results 2: " << results2 << std::endl;
  std::cout << "results 3: " << results3 << std::endl;

  return 0;
}

我看到的结果是:

results 0: []               
results 1: []               
results 2: []               
results 3: []  

1 个答案:

答案 0 :(得分:0)

使用cv::Mat打印cout时似乎有些怪异:您明确定义了指标的大小,但打印了空白结果。

您可以尝试不使用cv:Mat直接打印出值吗?您可以执行以下操作进行调试:

const auto* output = interpreter->typed_output_tensor<unsigned char>(0);
for (int i = 0; i < 40; ++i) {
  printf("%d, ", static_cast<int>(output[i]);
}