tensorflow C++ 调用PB模型文件参数问题

Question

问题遇到的现象和发生背景

最近在做NSFW识别，从github找到一个已经训练好的PB模型文件，使用python调用的，当尝试用C++调用的时候，发现参数不正确.通过python 生成日志，用TensorBoard查看日志时，发现模型输入参数是字符串？？字符串类型是json，在我理解力，明明应该是一张图片才对？！python 脚本里也是这样传的。
gitee地址: https://gitee.com/yyj8209/CVSample/tree/master/TensorFlow/inception_model
github源地址:https://github.com/kingroc711/CVSample/tree/master/TensorFlow/inception_model

问题相关代码，请勿粘贴截图

#include <iostream>
#include "tensorflow/core/public/session.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/framework/graph.pb.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/graph/default_device.h"
#include "tensorflow/core/graph/graph_def_builder.h"
#include "tensorflow/core/platform/logging.h"
#include "tensorflow/core/platform/types.h"
#include "tensorflow/core/public/session.h"

#include "tensorflow/cc/ops/standard_ops.h"
#include "tensorflow/cc/ops/image_ops.h"
//#include "eigen3/unsupported/Eigen/CXX11/Tensor"
//#include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
#include <vector>

//using namespace tensorflow;

using namespace std;
using namespace tensorflow;
using tensorflow::Tensor;
using tensorflow::Status;
using tensorflow::string;
using tensorflow::int32;



Status ReadTensorFromImageFile(string file_name, const int input_height,
                                 const int input_width,
                                 vector<Tensor>* out_tensors);

int main(int argc, char *argv[])
{
    SessionOptions sessionOptions;
    Session *session = NewSession(sessionOptions);
    string modelPath = "/opt/work/build_work/TensorFlow/inception_model/output_graph.pb";
    //tensorflow 官方模型
//    modelPath = "/opt/work/c_work/qt_work/tensorflow_cc_demo/model/classify_image_graph_def.pb";
    GraphDef graphDef;
    Status statud_load = ReadBinaryProto(Env::Default(), modelPath, &graphDef);
    if(statud_load.ok()) {
        cout << "load pb file success : " << modelPath << endl;
    }
    cout << "node size:" << graphDef.node_size() << endl;
    graphDef.node();

    if( session->Create(graphDef).ok() ) {
        cout << "success graph in session " << endl;
    }
    string image_path("/root/test.png");
//    image_path = "/opt/work/c_work/qt_work/tensorflow_cc_demo/model/cropped_panda.jpg";
    vector<Tensor> inputs;
    if(ReadTensorFromImageFile(image_path, 100, 100, &inputs).ok()) {
        cout << "image load success!" << endl;
        cout << inputs.size() << endl;
    }

    vector<Tensor> outputs;
    string input = "DecodeJpeg/contents:0";
    string output = "final_result:0";
    cout << inputs[0].DebugString() << endl;
    pair<string, Tensor> img(input,inputs[0]);
    vector<pair<string, tensorflow::Tensor>> runInputs = {
        {"DecodeJpeg/contents:0", inputs[0]},
    };

    Status status = session->Run(runInputs, {output}, {}, &outputs);
    cout << status << endl;
    if (!status.ok()) {
        cout << "run failed!" << endl;
    }
    cout << outputs.size() << endl;

    cout << "hello wordl" << endl;
    return 0;
}



Status ReadTensorFromImageFile(string file_name, const int input_height,
                                 const int input_width,
                                 vector<Tensor>* out_tensors) {
      auto root = Scope::NewRootScope();
      using namespace ops;

      auto file_reader = ops::ReadFile(root.WithOpName("file_reader"),file_name);
      const int wanted_channels = 1;
      Output image_reader;
      std::size_t found = file_name.find(".png");
      //判断文件格式
      if (found!=std::string::npos) {
          image_reader = DecodePng(root.WithOpName("png_reader"), file_reader,DecodePng::Channels(wanted_channels));
      }
      else {
          image_reader = DecodeJpeg(root.WithOpName("jpeg_reader"), file_reader,DecodeJpeg::Channels(wanted_channels));
      }
      // 下面几步是读取图片并处理
      auto float_caster =Cast(root.WithOpName("float_caster"), image_reader, DT_FLOAT);
      auto dims_expander = ExpandDims(root, float_caster, 0);
      auto resized = ResizeBilinear(root, dims_expander,Const(root.WithOpName("resize"), {input_height, input_width}));
      // Div(root.WithOpName(output_name), Sub(root, resized, {input_mean}),{input_std});
     Transpose(root.WithOpName("transpose"),resized,{0,2,1,3});

      GraphDef graph;
      root.ToGraphDef(&graph);

      unique_ptr<Session> session(NewSession(SessionOptions()));
      session->Create(graph);
      session->Run({}, {"transpose"}, {}, out_tensors);//Run，获取图片数据保存到Tensor中

      return Status::OK();
 }

运行结果及报错内容

Invalid argument: Expects arg[0] to be string but float is provided

我的解答思路和尝试过的方法

希望用C++调过PB模型的，给出一些建议。我没有接触过太多tensorflow，并不清楚我现在想法 对不对

我想要达到的结果

Answer 1

问题已解决，首先，传入参数是一张图片，类型确实是一个字符串,只不过类型是tstring，通过std::unique_ptrtensorflow::RandomAccessFile生成字符串。

tensorflow已经提供了针对C++预测图片提供了示例，示例地址:https://gitee.com/mirrors/tensorflow/blob/master/tensorflow/examples/label_image/main.cc
示例中ReadEntireFile函数就是专门处理把图片转为字符串数据的函数。
最后附上运行成功全部代码:

#include "tensorflow/core/public/session.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/framework/graph.pb.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/graph/default_device.h"
#include "tensorflow/core/graph/graph_def_builder.h"
#include "tensorflow/core/platform/logging.h"
#include "tensorflow/core/platform/types.h"
#include "tensorflow/core/public/session.h"
#include <vector>

using namespace std;
using namespace tensorflow;
using tensorflow::Tensor;
using tensorflow::Status;
using tensorflow::string;
using tensorflow::int32;

static Status ReadEntireFile(tensorflow::Env* env, const string& filename,
                                 Tensor* output);


int main(void){
    SessionOptions sessionOptions;
    Session *session = NewSession(sessionOptions);
    //pb文件路径
    string modelPath = "/opt/work/build_work/TensorFlow/inception_model/output_graph.pb";
    GraphDef graphDef;
    Status statud_load = ReadBinaryProto(Env::Default(), modelPath, &graphDef);
    if(statud_load.ok()) {
        cout << "load pb file success : " << modelPath << endl;
    }


    if( session->Create(graphDef).ok() ) {
        cout << "success graph in session " << endl;
    }
    vector<Tensor> outputs;
    string input = "DecodeJpeg/contents:0";
    string output = "final_result:0";
    Tensor input0(DT_STRING, TensorShape());
    //图片文件
    if(ReadEntireFile(tensorflow::Env::Default(), "/root/test.png", &input0).ok()) {
        cout << "图片读取成功！" << endl;
    }
    vector<pair<string, tensorflow::Tensor>> runInputs = {
        {"DecodeJpeg/contents:0", input0},
    };
    //预测
    Status status = session->Run(runInputs, {output}, {}, &outputs);
    cout << status << endl;
    if (!status.ok()) {
        cout << "run failed!" << endl;
    }
    //处理输出结果，模型输出结果就是一维数组，按照索引0，1，2，3，4分别对应porn。neutral、hentai、drawings、sexy
    Tensor scores;
    scores = outputs[0];
    tensorflow::TTypes<float>::Flat scores_flat = scores.flat<float>();
    //scores_flat.size() 数量是 5，打印每一个分类分数
    for(int i = 0; i < scores_flat.size(); i++) {
        cout << scores_flat(i) << endl;
    }
    return 0;
}


static Status ReadEntireFile(tensorflow::Env* env, const string& filename,
                             Tensor* output) {
  tensorflow::uint64 file_size = 0;
  TF_RETURN_IF_ERROR(env->GetFileSize(filename, &file_size));

  string contents;
  contents.resize(file_size);

  std::unique_ptr<tensorflow::RandomAccessFile> file;
  TF_RETURN_IF_ERROR(env->NewRandomAccessFile(filename, &file));

  tensorflow::StringPiece data;
  TF_RETURN_IF_ERROR(file->Read(0, file_size, &data, &(contents)[0]));
  if (data.size() != file_size) {
    return tensorflow::errors::DataLoss("Truncated read of '", filename,
                                        "' expected ", file_size, " got ",
                                        data.size());
  }
  output->scalar<tstring>()() = tstring(data);
  return Status::OK();
}
```c++

```