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             {"visualDebug", no_argument, 0, 'p'},
  47             {"help",      no_argument,       0,  'h' },
  48             {"output",    required_argument, 0,  'o' },
  49             {"visualize", optional_argument, 0,  'v' },
  50             {"fit",       optional_argument, 0,  'f' },
  51             {0,           0,                 0,  0 }
  52         };
  53
  54         int c = getopt_long(argc, argv, "dphv::f::o:", long_options, &option_index);
  55         if (c == -1)
  56             break;
  57
  58         switch (c) {
  59         case 'd':
  60             tracker.m_debug = true;
  61             break;
  62         case 'p':
  63             tracker.m_visual_debug = true;
  64             visualize_delay = 500;
  65             break;
  66         case 'h':
  67             std::cerr << "Usage: \n"
  68                       << argv[0] << " [options]\n"
  69                       << argv[0] << " [options] <directory>\n"
  70                       << argv[0] << " [options] <path/to/region.txt or groundtruth.txt> <path/to/images.txt> [path/to/output.txt]\n"
  71                       << "Options:\n"
  72                       << " --visualize | -v[delay_ms]\n"
  73                       << " --output    | -o <output.txt>\n"
  74                       << " --debug     | -d\n"
  75                       << " --fit       | -f[W[xH]]\n";
  76             exit(0);
  77             break;
  78         case 'o':
  79             output = optarg;
  80             break;
  81         case 'v':
  82             visualize_delay = optarg ? atol(optarg) : 1;
  83             break;
  84         case 'f':
  85             if (!optarg) {
  86                 fit_size_x = fit_size_y = 0;
  87             } else {
  88                 char tail;
  89                 if (sscanf(optarg, "%d%c", &fit_size_x, &tail) == 1) {
  90                     fit_size_y = fit_size_x;
  91                 } else if (sscanf(optarg, "%dx%d%c", &fit_size_x, &fit_size_y, &tail) != 2) {
  92                     fprintf(stderr, "Cannot parse -f argument: %s\n", optarg);
  93                     return 1;
  94                 }
  95             }
  96             break;
  97         }
  98     }
  99
 100     switch (argc - optind) {
 101     case 1:
 102         if (chdir(argv[optind]) == -1) {
 103             perror(argv[optind]);
 104             exit(1);
 105         }
 106         // Fall through
 107     case 0:
 108         region = access("groundtruth.txt", F_OK) == 0 ? "groundtruth.txt" : "region.txt";
 109         images = "images.txt";
 110         if (output.empty())
 111             output = "output.txt";
 112         break;
 113     case 2:
 114         // Fall through
 115     case 3:
 116         region = std::string(argv[optind + 0]);
 117         images = std::string(argv[optind + 1]);
 118         if (output.empty()) {
 119             if ((argc - optind) == 3)
 120                 output = std::string(argv[optind + 2]);
 121             else
 122                 output = std::string(dirname(argv[optind + 0])) + "/output.txt";
 123         }
 124         break;
 125     default:
 126         std::cerr << "Too many arguments\n";
 127         return 1;
 128     }
 129     VOT vot_io(region, images, output);
 130
 131     // if groundtruth.txt is used use intersection over union (IOU) to calculate tracker accuracy
 132     std::ifstream groundtruth_stream;
 133     if (region.compare("groundtruth.txt") == 0) {
 134         groundtruth_stream.open(region.c_str());
 135         std::string line;
 136         std::getline(groundtruth_stream, line);
 137     }
 138
 139     cv::Mat image;
 140
 141     //img = firts frame, initPos = initial position in the first frame
 142     cv::Rect init_rect = vot_io.getInitRectangle();
 143     vot_io.outputBoundingBox(init_rect);
 144     vot_io.getNextImage(image);
 145
 146     tracker.init(image, init_rect, fit_size_x, fit_size_y);
 147
 148     BBox_c bb;
 149     cv::Rect bb_rect;
 150     double avg_time = 0., sum_accuracy = 0.;
 151     int frames = 0;
 152
 153     std::cout << std::fixed << std::setprecision(2);
 154
 155     while (vot_io.getNextImage(image) == 1){
 156         double time_profile_counter = cv::getCPUTickCount();
 157         tracker.track(image);
 158         time_profile_counter = cv::getCPUTickCount() - time_profile_counter;
 159          std::cout << "  -> speed : " <<  time_profile_counter/((double)cvGetTickFrequency()*1000) << "ms per frame, "
 160                       "response : " << tracker.getFilterResponse();
 161         avg_time += time_profile_counter/((double)cvGetTickFrequency()*1000);
 162         frames++;
 163
 164         bb = tracker.getBBox();
 165         bb_rect = cv::Rect(bb.cx - bb.w/2., bb.cy - bb.h/2., bb.w, bb.h);
 166         vot_io.outputBoundingBox(bb_rect);
 167
 168         if (groundtruth_stream.is_open()) {
 169             std::string line;
 170             std::getline(groundtruth_stream, line);
 171
 172             cv::Rect groundtruthRect;
 173             double accuracy = calcAccuracy(line, bb_rect, groundtruthRect);
 174             if (visualize_delay >= 0)
 175                 cv::rectangle(image, groundtruthRect, CV_RGB(255, 0,0), 1);
 176             std::cout << ", accuracy: " << accuracy;
 177             sum_accuracy += accuracy;
 178         }
 179
 180         std::cout << std::endl;
 181
 182         if (visualize_delay >= 0) {
 183             cv::Point pt(bb.cx, bb.cy);
 184             cv::Size size(bb.w, bb.h);
 185             cv::RotatedRect rotatedRectangle(pt, size, bb.a);
 186
 187             cv::Point2f vertices[4];
 188             rotatedRectangle.points(vertices);
 189
 190             for (int i = 0; i < 4; i++)
 191                 cv::line(image, vertices[i], vertices[(i + 1) % 4], cv::Scalar(0, 255, 0), 2);
 192             cv::imshow("KCF output", image);
 193             int ret = cv::waitKey(visualize_delay);
 194             if (visualize_delay > 0 && ret != -1 && ret < 128)
 195                 break;
 196         }
 197
 198 //        std::stringstream s;
 199 //        std::string ss;
 200 //        int countTmp = frames;
 201 //        s << "imgs" << "/img" << (countTmp/10000);
 202 //        countTmp = countTmp%10000;
 203 //        s << (countTmp/1000);
 204 //        countTmp = countTmp%1000;
 205 //        s << (countTmp/100);
 206 //        countTmp = countTmp%100;
 207 //        s << (countTmp/10);
 208 //        countTmp = countTmp%10;
 209 //        s << (countTmp);
 210 //        s << ".jpg";
 211 //        s >> ss;
 212 //        //set image output parameters
 213 //        std::vector<int> compression_params;
 214 //        compression_params.push_back(CV_IMWRITE_JPEG_QUALITY);
 215 //        compression_params.push_back(90);
 216 //        cv::imwrite(ss.c_str(), image, compression_params);
 217     }
 218
 219     std::cout << "Average processing speed: " << avg_time / frames << "ms (" << 1. / (avg_time / frames) * 1000 << " fps)";
 220     if (groundtruth_stream.is_open()) {
 221         std::cout << "; Average accuracy: " << sum_accuracy/frames << std::endl;
 222         groundtruth_stream.close();
 223     }
 224     std::cout << std::endl;
 225
 226     return EXIT_SUCCESS;
 227 }