main_vot.cpp

   1 #include <stdlib.h>
   2 #include <getopt.h>
   3 #include <libgen.h>
   4 #include <unistd.h>
   5
   6 #include "kcf.h"
   7 #include "vot.hpp"
   8
   9 double calcAccuracy(std::string line, cv::Rect bb_rect, cv::Rect &groundtruth_rect)
  10 {
  11     std::vector<float> numbers;
  12     std::istringstream s( line );
  13     float x;
  14     char ch;
  15
  16     while (s >> x){
  17         numbers.push_back(x);
  18         s >> ch;
  19     }
  20     double x1 = std::min(numbers[0], std::min(numbers[2], std::min(numbers[4], numbers[6])));
  21     double x2 = std::max(numbers[0], std::max(numbers[2], std::max(numbers[4], numbers[6])));
  22     double y1 = std::min(numbers[1], std::min(numbers[3], std::min(numbers[5], numbers[7])));
  23     double y2 = std::max(numbers[1], std::max(numbers[3], std::max(numbers[5], numbers[7])));
  24
  25     groundtruth_rect = cv::Rect(x1, y1, x2-x1, y2-y1);
  26
  27     double rects_intersection = (groundtruth_rect & bb_rect).area();
  28     double rects_union = (groundtruth_rect | bb_rect).area();
  29     double accuracy = rects_intersection/rects_union;
  30
  31     return accuracy;
  32 }
  33
  34 int main(int argc, char *argv[])
  35 {
  36     //load region, images and prepare for output
  37     std::string region, images, output;
  38     int visualize_delay = -1, fit_size_x = -1, fit_size_y = -1;
  39     KCF_Tracker tracker;
  40
  41     while (1) {
  42         int option_index = 0;
  43         static struct option long_options[] = {
  44             {"debug",     no_argument,       0,  'd' },
  45             {"visualDebug",     no_argument,       0,  'p' },
  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, "dphv::f::o:",
  54                         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                       << " --visualDebug | -p\n"
  76                       << " --fit       | -f[WxH]\n";
  77             exit(0);
  78             break;
  79         case 'o':
  80             output = optarg;
  81             break;
  82         case 'v':
  83             visualize_delay = optarg ? atol(optarg) : 1;
  84             break;
  85         case 'f':
  86             std::string sizes = optarg ? optarg : "128x128";
  87             std::string delimiter = "x";
  88             size_t pos = sizes.find(delimiter);
  89             std::string first_argument = sizes.substr(0, pos);
  90             sizes.erase(0, pos + delimiter.length());
  91
  92             fit_size_x = stol(first_argument);
  93             fit_size_y = stol(sizes);
  94             break;
  95         }
  96     }
  97
  98     switch (argc - optind) {
  99     case 1:
 100         if (chdir(argv[optind]) == -1) {
 101             perror(argv[optind]);
 102             exit(1);
 103         }
 104         // Fall through
 105     case 0:
 106         region = access("groundtruth.txt", F_OK) == 0 ? "groundtruth.txt" : "region.txt";
 107         images = "images.txt";
 108         if (output.empty())
 109             output = "output.txt";
 110         break;
 111     case 2:
 112         // Fall through
 113     case 3:
 114         region = std::string(argv[optind + 0]);
 115         images = std::string(argv[optind + 1]);
 116         if (output.empty()) {
 117             if ((argc - optind) == 3)
 118                 output = std::string(argv[optind + 2]);
 119             else
 120                 output = std::string(dirname(argv[optind + 0])) + "/output.txt";
 121         }
 122         break;
 123     default:
 124         std::cerr << "Too many arguments\n";
 125         return 1;
 126     }
 127     VOT vot_io(region, images, output);
 128
 129     // if groundtruth.txt is used use intersection over union (IOU) to calculate tracker accuracy
 130     std::ifstream groundtruth_stream;
 131     if (region.compare("groundtruth.txt") == 0) {
 132         std::cout << region << std::endl;
 133         groundtruth_stream.open(region.c_str());
 134         std::string line;
 135         std::getline(groundtruth_stream, line);
 136     }
 137
 138     cv::Mat image;
 139
 140     //img = firts frame, initPos = initial position in the first frame
 141     cv::Rect init_rect = vot_io.getInitRectangle();
 142     vot_io.outputBoundingBox(init_rect);
 143     vot_io.getNextImage(image);
 144
 145     tracker.init(image, init_rect, fit_size_x, fit_size_y);
 146
 147     BBox_c bb;
 148     cv::Rect bb_rect;
 149     double avg_time = 0., sum_accuracy = 0.;
 150     int frames = 0;
 151     while (vot_io.getNextImage(image) == 1){
 152         double time_profile_counter = cv::getCPUTickCount();
 153         tracker.track(image);
 154         time_profile_counter = cv::getCPUTickCount() - time_profile_counter;
 155          std::cout << "  -> speed : " <<  time_profile_counter/((double)cvGetTickFrequency()*1000) << "ms. per frame";
 156         avg_time += time_profile_counter/((double)cvGetTickFrequency()*1000);
 157         frames++;
 158
 159         bb = tracker.getBBox();
 160         bb_rect = cv::Rect(bb.cx - bb.w/2., bb.cy - bb.h/2., bb.w, bb.h);
 161         vot_io.outputBoundingBox(bb_rect);
 162
 163         if (groundtruth_stream.is_open()) {
 164             std::string line;
 165             std::getline(groundtruth_stream, line);
 166
 167             cv::Rect groundtruthRect;
 168             double accuracy = calcAccuracy(line, bb_rect, groundtruthRect);
 169             if (visualize_delay >= 0)
 170                 cv::rectangle(image, groundtruthRect, CV_RGB(255, 0,0), 1);
 171             std::cout << ", accuracy: " << accuracy;
 172             sum_accuracy += accuracy;
 173         }
 174
 175         std::cout << std::endl;
 176
 177         if (visualize_delay >= 0) {
 178             cv::Point pt(bb.cx, bb.cy);
 179             cv::Size size(bb.w, bb.h);
 180             cv::RotatedRect rotatedRectangle(pt,size, bb.a);
 181
 182             cv::Point2f vertices[4];
 183             rotatedRectangle.points(vertices);
 184
 185             for (int i = 0; i < 4; i++)
 186                 cv::line(image, vertices[i], vertices[(i+1)%4], cv::Scalar(0,255,0), 2);
 187 //             cv::rectangle(image, cv::Rect(bb.cx - bb.w/2., bb.cy - bb.h/2., bb.w, bb.h), CV_RGB(0,255,0), 2);
 188             std::string angle = std::to_string (bb.a);
 189             angle.erase ( angle.find_last_not_of('0') + 1, std::string::npos );
 190             angle.erase ( angle.find_last_not_of('.') + 1, std::string::npos );
 191             cv::putText(image, "Frame: " + std::to_string(frames) + " " + angle + " angle", cv::Point(0, image.rows-1), cv::FONT_HERSHEY_SIMPLEX, 0.7, cv::Scalar(0,255,0),2,cv::LINE_AA);
 192             cv::imshow("output", image);
 193             int ret = cv::waitKey(visualize_delay);
 194             if (visualize_delay > 0 && ret != -1 && ret != 255)
 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)" << std::endl;
 220     if (groundtruth_stream.is_open()) {
 221         std::cout << "Average accuracy: " << sum_accuracy/frames << std::endl;
 222         groundtruth_stream.close();
 223     }
 224
 225     return EXIT_SUCCESS;
 226 }