FastDeploy 安装教程

【2023-2-22】FastDeploy 安装编译教程

该测试 FastDeploy CPU版本。

1. fastDeploy库编译

1.1 官方预编译库下载

预编译库下载安装

1.2 自定义CPU版本库编译

• 官方编译FastDeploy教程

CMakeGUI + VS 2019 IDE编译FastDeploy

• 本人编译教程

1. CMAKE_CONFIGURATION_TYPES 属性设置为Release

2. 请不要勾选WITH_GPU和ENABLE_TRT_BACKEND

3. 开启ENABLE_PADDLE_BACKEND ENABLE_OPENVINO_BACKEND ENABLE_VISION

4. 指定CMAKE_INSTALL_PREFIX 安装路径

5. 生成fastdeploy.sln解决方案文件选择Release版本，生成编译，点击"INSTALL"->右键点击"生成"将编译好的SDK安装到先前指定的目录步骤⑤。
   
   

1.3 自定义GPU版本库编译

• 官方编译FastDeploy教程

CMakeGUI + VS 2019 IDE编译FastDeploy

• 本人编译教程

1. CMAKE_CONFIGURATION_TYPES 属性设置为Release

2. 勾选WITH_GPU和ENABLE_TRT_BACKEND

3. 开启ENABLE_ORT_BACKEND ENABLE_PADDLE_BACKEND ENABLE_TRT_BACKEND ENABLE_OPENVINO_BACKEND ENABLE_VISION ENABLE_TEXT

4. 设置CUDA TensorRT路径

CUDA_DIRECTORY            : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.2
TRT_DRECTORY              : D:/Program Files/TensorRT-8.4.3.1

5. 指定CMAKE_INSTALL_PREFIX 安装路径

6. 生成fastdeploy.sln解决方案文件选择Release版本，生成编译，点击"INSTALL"->右键点击"生成"将编译好的SDK安装到先前指定的目录步骤⑤。
   
   

1.4 样例测试

picodet_l_320_coco_lcnet 模型下载

/*
 * @Description: 
 * @Version: 0.0.1
 * @Author: chccc
 * @Date: 2023-02-19 00:19:22
 * @LastEditors: chccc
 * @LastEditTime: 2023-02-20 17:23:21
 * @FilePath: \cpp\infer.cpp
 */
#include 
#include "fastdeploy/vision.h"

void PicoDetGpuInfer(const std::string& model_dir, const std::string& image_file)
{

	std::string model_file = "picodet_l_320_coco_lcnet/model.pdmodel";
	std::string params_file ="picodet_l_320_coco_lcnet/model.pdiparams";
	std::string config_file ="picodet_l_320_coco_lcnet/infer_cfg.yml";;

	auto option = fastdeploy::RuntimeOption();
	option.UseGpu(0);
	fastdeploy::vision::detection::PicoDet model = fastdeploy::vision::detection::PicoDet(model_file, params_file, config_file, option);
	std::cout<<typeid(model).name()<<std::endl;
	if (!model.Initialized()) {
		printf("[%s][%d] Error: fastdeploy::vision::detection::PicoDet initialized failed !\n", __func__, __LINE__);
		return;
	}

	auto im = cv::imread(image_file);
	auto im_bak = im.clone();

	auto start = std::chrono::system_clock::now();

	fastdeploy::vision::DetectionResult result;
	if (!model.Predict(&im, &result)) {
		printf("[%s][%d] Error: Failed to predict !\n", __func__, __LINE__);
		return;
	}
	//std::cout << res.Str() << std::endl;

	auto end = std::chrono::system_clock::now();
	auto duration = std::chrono::duration_cast<std::chrono::microseconds>(end - start);
	double costTime = double(duration.count()) * std::chrono::microseconds::period::num / std::chrono::microseconds::period::den;
	//printf("[%s][%d] model.Predict success, cost time: %lf s \n", __func__, __LINE__, costTime);

	std::cout << "==========================================================================" << std::endl;

    float score_thereshold = 0.8;
	int line_size = 2;
	float font_size = 1;
	auto vis_image = fastdeploy::vision::Visualize::VisDetection(im_bak, result, score_thereshold, line_size, font_size);
	std::string vis_image_path = "./images/vis_result.jpg";
	cv::imwrite(vis_image_path, vis_image);
	printf("[%s][%d] Visualized result saved in %s !\n", __func__, __LINE__, vis_image_path.c_str());
}

int main()
{
	std::cout << "=============== TEST ===============" << std::endl;

	std::string model_dir = "./models/";;
	std::string image_file = "./images/1.jpg";

	//计时
	auto start = std::chrono::system_clock::now();

	PicoDetGpuInfer(model_dir, image_file);

	//计时
	auto end = std::chrono::system_clock::now();
	auto duration = std::chrono::duration_cast<std::chrono::microseconds>(end - start);
	double costTime = double(duration.count()) * std::chrono::microseconds::period::num / std::chrono::microseconds::period::den;

	printf("[%s][%d] Model infer success, cost time: %lf s \n", __func__, __LINE__, costTime);

	std::cout << "=============== Finished ===============" << std::endl;
}