YOLO视频检测的流程细节

YOLO检测的三大步骤

1. 载入cfg
2. 载入weights
3. 网络预测

C:\darknet-master\src\art.c  

void demo_art(char *cfgfile, char *weightfile, int cam_index)
{
#ifdef OPENCV
    network net = parse_network_cfg(cfgfile);//载入cfg
    if(weightfile){//载入weights
        load_weights(&net, weightfile);
    }
    set_batch_network(&net, 1);

    srand(2222222);
    cap_cv * cap;

    cap = get_capture_webcam(cam_index);

    char *window = "ArtJudgementBot9000!!!";
    if(!cap) error("Couldn't connect to webcam.\n");
    create_window_cv(window, 0, 512, 512);
    int i;
    int idx[] = {37, 401, 434};
    int n = sizeof(idx)/sizeof(idx[0]);

    while(1){
        image in = get_image_from_stream_cpp(cap);
        image in_s = resize_image(in, net.w, net.h);
        show_image(in, window);

        float *p = network_predict(net, in_s.data);//网络预测

        printf("\033[2J");
        printf("\033[1;1H");

        float score = 0;
        for(i = 0; i < n; ++i){
            float s = p[idx[i]];
            if (s > score) score = s;
        }
        score = score;
        printf("I APPRECIATE THIS ARTWORK: %10.7f%%\n", score*100);
        printf("[");
	int upper = 30;
        for(i = 0; i < upper; ++i){
            printf("%c", ((i+.5) < score*upper) ? 219 : ' ');
        }
        printf("]\n");

        free_image(in_s);
        free_image(in);

        wait_key_cv(1);
    }
#endif
}

从图中代码可以看出步骤1,2仅进行的1次，步骤3在一直循环

基于src/yolo_v2_class.cpp的lib扩展的独立工程中这三个步骤的分布所在。

1. 载入cfg
2. 载入weights
   位于构造函数中

class Detector {
    std::shared_ptr<void> detector_gpu_ptr;
    std::deque<std::vector<bbox_t>> prev_bbox_vec_deque;
public:
    const int cur_gpu_id;
    float nms = .4;
    bool wait_stream;

    LIB_API Detector(std::string cfg_filename, std::string weight_filename, int gpu_id = 0);

3. 网络预测
   在detect函数里

LIB_API std::vector<bbox_t> Detector::detect(image_t img, float thresh, bool use_mean)
{
    detector_gpu_t &detector_gpu = *static_cast<detector_gpu_t *>(detector_gpu_ptr.get());
    network &net = detector_gpu.net;
#ifdef GPU
    int old_gpu_index;
    cudaGetDevice(&old_gpu_index);
    if(cur_gpu_id != old_gpu_index)
        cudaSetDevice(net.gpu_index);

    net.wait_stream = wait_stream;    // 1 - wait CUDA-stream, 0 - not to wait
#endif
    //std::cout << "net.gpu_index = " << net.gpu_index << std::endl;

    image im;
    im.c = img.c;
    im.data = img.data;
    im.h = img.h;
    im.w = img.w;

    image sized;

    if (net.w == im.w && net.h == im.h) {
        sized = make_image(im.w, im.h, im.c);
        memcpy(sized.data, im.data, im.w*im.h*im.c * sizeof(float));
    }
    else
        sized = resize_image(im, net.w, net.h);

    layer l = net.layers[net.n - 1];

    float *X = sized.data;

    float *prediction = network_predict(net, X);