tensorflow lite example label_image 分析【二】

接上文

3.代码分析

main函数首先将入参写入参数结构体

Settings s;

struct Settings {
  bool verbose = false;
  bool accel = false;
  bool input_floating = false;
  bool profiling = false;
  int loop_count = 1;
  float input_mean = 127.5f;
  float input_std = 127.5f;
  string model_name = "./mobilenet_quant_v1_224.tflite";
  string input_bmp_name = "./grace_hopper.bmp";
  string labels_file_name = "./labels.txt";
  string input_layer_type = "uint8_t";
  int number_of_threads = 4;

};

运行执行函数 RunInference(&s);

【背景知识：https://blog.csdn.net/jILRvRTrc/article/details/80553561】

RunInference();

1）建立模型model = tflite::FlatBufferModel::BuildFromFile(s->model_name.c_str());

2）建立OpResolver 用于指向每个node的操作函数 tflite::ops::builtin::B