Function main

lesson7-Detection/src/main.cpp:75–121

}

int main()
{
	//float tt[20] = { 0.4928,  0.2981,  0.1731,  0.4038, 17.0000,0.3317,  0.7163,  0.5673,  0.2548,  0.0000,0.8714,
	//0.8125,  0.1370,  0.2308, 14.0000,0.6250,  0.6202,  0.0144,  0.0240, 14.0000 };
	//auto targets = torch::from_blob(tt, { 4,5 }, torch::TensorOptions(torch::kFloat32));

	//float anchor[12] = { 10,14,  23,27,  37,58,  81,82,  135,169,  344,319 };
	//auto anchors_ = torch::from_blob(anchor, { 6,2 }, torch::TensorOptions(torch::kFloat32)).to(torch::Device(torch::kCUDA));
	//int image_size[2] = { 416, 416 };

	//bool normalize = false;
	//auto critia1 = YOLOLossImpl(anchors_, 20, image_size, 0.01, torch::Device(torch::kCUDA), normalize);

	//auto output = torch::ones({ 1,75,13,13 }, torch::kFloat).to(torch::Device(torch::kCUDA));
	////auto tmp = torch::tensor({ 0.2, 0.2, 0.3, 0.3, 1.0 }, torch::kFloat).unsqueeze(0).to(torch::Device(torch::kCUDA));
	////std::cout << tmp.sizes();
	//std::vector<torch::Tensor> targets_vec = { targets };
	//auto loss_numpos1 = critia1.forward(output, targets_vec);
	//std::cout << loss_numpos1[0];



	cv::Mat image = cv::imread("D:\\AllentFiles\\data\\dataset4work\\detection\\val\\images\\2007_005331.jpg");
	Detector detector;
	detector.Initialize(-1, 416, 416, "D:\\AllentFiles\\data\\dataset4work\\detection\\name.txt");
	//detector.Train("D:\\AllentFiles\\data\\dataset4work\\detection", ".jpg", 30,
	//	4, 0.001, "weights/detector.pt", "weights/yolo4_tiny.pt");

	detector.LoadWeight("weights/detector.pt");


	auto model = YoloBody_tiny(3, 80);
	torch::load(model, "weights/yolo4_tiny.pt");
	model->to(torch::kCUDA);
	Predict(model, image, true, 0.1, 0.3);

	detector.Predict(image, true, 0.1);
	int64 start = cv::getTickCount();
	int loops = 10;
	for (int i = 0; i < loops; i++) {
		detector.Predict(image, false);
	}
	double duration = (cv::getTickCount() - start) / cv::getTickFrequency();
	std::cout << duration/ loops <<" s per prediction" << std::endl;

	return 0;
}