tflite C++ API 部署分类模型

最新推荐文章于 2023-08-11 15:27:04 发布

烂桃先生之前的账号

最新推荐文章于 2023-08-11 15:27:04 发布

阅读量4.5k

点赞数 1

分类专栏：深度学习移动端部署

本文链接：https://blog.youkuaiyun.com/weixin_42464187/article/details/103362255

版权

本文档详细介绍了如何使用TensorFlowLite C++ API来部署分类模型，包括编写CMakeLists.txt，设置配置，引入所需头文件，以及主要函数的介绍，参照官方GitHub仓库提供的示例。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

TensorFlowLite C++ API 分类模型

文章目录

TensorFlowLite C++ API 分类模型

1、编写CMakeLists.txt

# 添加
set(TFLITE_LIBS /home/surui/Downloads/software/tensorflow-master/tensorflow/lite/tools/make/gen/linux_x86_64/lib/libtensorflow-lite.a)
include_directories(/home/surui/Downloads/software/tensorflow-master/)
include_directories(/home/surui/Downloads/software/tensorflow-master/tensorflow/lite/tools/make/downloads/flatbuffers/include)
include_directories(/home/surui/Downloads/software/tensorflow-master/tensorflow/lite/tools/make/downloads/absl)

2、定义Settings

namespace tflite {
namespace label_image {

struct Settings {
  bool verbose = false;
  bool accel = false;
  bool old_accel = false;
  bool input_floating = false;
  bool profiling = false;
  bool allow_fp16 = false;
  bool gl_backend = false;
  int loop_count = 1;
  float input_mean = 127.5f;
  float input_std = 127.5f;
  string model_name = "./mobilenet_quant_v1_224.tflite";
  tflite::FlatBufferModel* model;
  string input_bmp_name = "./grace_hopper.bmp";
  string labels_file_name = "./labels.txt";
  string input_layer_type = "uint8_t";
  int number_of_threads = 4;
  int number_of_results = 5;
  int max_profiling_buffer_entries = 1024;
  int number_of_warmup_runs = 2;
};

}  // namespace label_image
}  // namespace tflite

3、引入头文件

#include "tensorflow/lite/examples/label_image/label_image.h"

#include <f

最低0.47元/天解锁文章