main_vot.cpp

   1 #include <stdlib.h>
   2 #include <getopt.h>
   3 #include <libgen.h>
   4 #include <unistd.h>
   5 #include <iomanip>
   6
   7 #include "kcf.h"
   8 #include "vot.hpp"
   9
  10 double calcAccuracy(std::string line, cv::Rect bb_rect, cv::Rect &groundtruth_rect)
  11 {
  12     std::vector<float> numbers;
  13     std::istringstream s(line);
  14     float x;
  15     char ch;
  16
  17     while (s >> x) {
  18         numbers.push_back(x);
  19         s >> ch;
  20     }
  21     double x1 = std::min(numbers[0], std::min(numbers[2], std::min(numbers[4], numbers[6])));
  22     double x2 = std::max(numbers[0], std::max(numbers[2], std::max(numbers[4], numbers[6])));
  23     double y1 = std::min(numbers[1], std::min(numbers[3], std::min(numbers[5], numbers[7])));
  24     double y2 = std::max(numbers[1], std::max(numbers[3], std::max(numbers[5], numbers[7])));
  25
  26     groundtruth_rect = cv::Rect(x1, y1, x2 - x1, y2 - y1);
  27
  28     double rects_intersection = (groundtruth_rect & bb_rect).area();
  29     double rects_union = (groundtruth_rect | bb_rect).area();
  30     double accuracy = rects_intersection / rects_union;
  31
  32     return accuracy;
  33 }
  34
  35 int main(int argc, char *argv[])
  36 {
  37     //load region, images and prepare for output
  38     std::string region, images, output;
  39     int visualize_delay = -1, fit_size_x = -1, fit_size_y = -1;
  40     KCF_Tracker tracker;
  41
  42     while (1) {
  43         int option_index = 0;
  44         static struct option long_options[] = {
  45             {"debug",     no_argument,       0,  'd' },
  46             {"help",      no_argument,       0,  'h' },
  47             {"output",    required_argument, 0,  'o' },
  48             {"visualize", optional_argument, 0,  'v' },
  49             {"fit",       optional_argument, 0,  'f' },
  50             {0,           0,                 0,  0 }
  51         };
  52
  53         int c = getopt_long(argc, argv, "dhv::f::o:", long_options, &option_index);
  54         if (c == -1)
  55             break;
  56
  57         switch (c) {
  58         case 'd':
  59             tracker.m_debug = true;
  60             break;
  61         case 'h':
  62             std::cerr << "Usage: \n"
  63                       << argv[0] << " [options]\n"
  64                       << argv[0] << " [options] <directory>\n"
  65                       << argv[0] << " [options] <path/to/region.txt or groundtruth.txt> <path/to/images.txt> [path/to/output.txt]\n"
  66                       << "Options:\n"
  67                       << " --visualize | -v[delay_ms]\n"
  68                       << " --output    | -o <output.txt>\n"
  69                       << " --debug     | -d\n"
  70                       << " --fit       | -f[WxH]\n";
  71             exit(0);
  72             break;
  73         case 'o':
  74             output = optarg;
  75             break;
  76         case 'v':
  77             visualize_delay = optarg ? atol(optarg) : 1;
  78             break;
  79         case 'f':
  80             if (!optarg) {
  81                 fit_size_x = fit_size_y = 128;
  82             } else {
  83                 char tail;
  84                 if (sscanf(optarg, "%d%c", &fit_size_x, &tail) == 1) {
  85                     fit_size_y = fit_size_x;
  86                 } else if (sscanf(optarg, "%dx%d%c", &fit_size_x, &fit_size_y, &tail) != 2) {
  87                     fprintf(stderr, "Cannot parse -f argument: %s\n", optarg);
  88                     return 1;
  89                 }
  90             }
  91             int min_size = 2 * tracker.p_cell_size;
  92             if (fit_size_x <  min_size || fit_size_x < min_size) {
  93                 fprintf(stderr, "Fit size %dx%d too small. Minimum is %dx%d.\n",
  94                         fit_size_x, fit_size_y, min_size, min_size);
  95                 return 1;
  96             }
  97             break;
  98         }
  99     }
 100
 101     switch (argc - optind) {
 102     case 1:
 103         if (chdir(argv[optind]) == -1) {
 104             perror(argv[optind]);
 105             exit(1);
 106         }
 107         // Fall through
 108     case 0:
 109         region = access("groundtruth.txt", F_OK) == 0 ? "groundtruth.txt" : "region.txt";
 110         images = "images.txt";
 111         if (output.empty())
 112             output = "output.txt";
 113         break;
 114     case 2:
 115         // Fall through
 116     case 3:
 117         region = std::string(argv[optind + 0]);
 118         images = std::string(argv[optind + 1]);
 119         if (output.empty()) {
 120             if ((argc - optind) == 3)
 121                 output = std::string(argv[optind + 2]);
 122             else
 123                 output = std::string(dirname(argv[optind + 0])) + "/output.txt";
 124         }
 125         break;
 126     default:
 127         std::cerr << "Too many arguments\n";
 128         return 1;
 129     }
 130     VOT vot_io(region, images, output);
 131
 132     // if groundtruth.txt is used use intersection over union (IOU) to calculate tracker accuracy
 133     std::ifstream groundtruth_stream;
 134     if (region.compare("groundtruth.txt") == 0) {
 135         groundtruth_stream.open(region.c_str());
 136         std::string line;
 137         std::getline(groundtruth_stream, line);
 138     }
 139
 140     cv::Mat image;
 141
 142     //img = firts frame, initPos = initial position in the first frame
 143     cv::Rect init_rect = vot_io.getInitRectangle();
 144     vot_io.outputBoundingBox(init_rect);
 145     vot_io.getNextImage(image);
 146
 147     tracker.init(image, init_rect, fit_size_x, fit_size_y);
 148
 149     BBox_c bb;
 150     cv::Rect bb_rect;
 151     double avg_time = 0., sum_accuracy = 0.;
 152     int frames = 0;
 153
 154     std::cout << std::fixed << std::setprecision(2);
 155
 156     while (vot_io.getNextImage(image) == 1){
 157         double time_profile_counter = cv::getCPUTickCount();
 158         tracker.track(image);
 159         time_profile_counter = cv::getCPUTickCount() - time_profile_counter;
 160          std::cout << "  -> speed : " <<  time_profile_counter/((double)cvGetTickFrequency()*1000) << "ms per frame, "
 161                       "response : " << tracker.getFilterResponse();
 162         avg_time += time_profile_counter/((double)cvGetTickFrequency()*1000);
 163         frames++;
 164
 165         bb = tracker.getBBox();
 166         bb_rect = cv::Rect(bb.cx - bb.w/2., bb.cy - bb.h/2., bb.w, bb.h);
 167         vot_io.outputBoundingBox(bb_rect);
 168
 169         if (groundtruth_stream.is_open()) {
 170             std::string line;
 171             std::getline(groundtruth_stream, line);
 172
 173             cv::Rect groundtruthRect;
 174             double accuracy = calcAccuracy(line, bb_rect, groundtruthRect);
 175             if (visualize_delay >= 0)
 176                 cv::rectangle(image, groundtruthRect, CV_RGB(255, 0,0), 1);
 177             std::cout << ", accuracy: " << accuracy;
 178             sum_accuracy += accuracy;
 179         }
 180
 181         std::cout << std::endl;
 182
 183         if (visualize_delay >= 0) {
 184             cv::rectangle(image, bb_rect, CV_RGB(0,255,0), 2);
 185             cv::imshow("output", image);
 186             int ret = cv::waitKey(visualize_delay);
 187             if (visualize_delay > 0 && ret != -1 && ret < 128)
 188                 break;
 189         }
 190
 191 //        std::stringstream s;
 192 //        std::string ss;
 193 //        int countTmp = frames;
 194 //        s << "imgs" << "/img" << (countTmp/10000);
 195 //        countTmp = countTmp%10000;
 196 //        s << (countTmp/1000);
 197 //        countTmp = countTmp%1000;
 198 //        s << (countTmp/100);
 199 //        countTmp = countTmp%100;
 200 //        s << (countTmp/10);
 201 //        countTmp = countTmp%10;
 202 //        s << (countTmp);
 203 //        s << ".jpg";
 204 //        s >> ss;
 205 //        //set image output parameters
 206 //        std::vector<int> compression_params;
 207 //        compression_params.push_back(CV_IMWRITE_JPEG_QUALITY);
 208 //        compression_params.push_back(90);
 209 //        cv::imwrite(ss.c_str(), image, compression_params);
 210     }
 211
 212     std::cout << "Average processing speed: " << avg_time / frames << "ms (" << 1. / (avg_time / frames) * 1000 << " fps)";
 213     if (groundtruth_stream.is_open()) {
 214         std::cout << "; Average accuracy: " << sum_accuracy/frames << std::endl;
 215         groundtruth_stream.close();
 216     }
 217     std::cout << std::endl;
 218
 219     return EXIT_SUCCESS;
 220 }