#include "MultiTracker.h"
-#include "Metrics.h"
#include <algorithm>
#include "hungarian.h"
#include "Logger.h"
#include "Utils.h"
+#include <cstdlib>
+#include <stdexcept>
using namespace suanzi;
using namespace cv;
static const std::string TAG = "MultiTracker";
static const cv::Size PREFERRED_SIZE = Size(64, 128);
-
static const double MaxCost = 100000;
-static const int MaxPath = 5;
+static const double ProbThreshold = 0.05;
+static const int MaxTrackers = 100;
MultiTracker::MultiTracker(EngineWPtr e)
: engine(e)
{
LOG_DEBUG(TAG, "init - loading model.pkl");
- predictor = PredictorWrapper::create("./python", "./python/model.pkl");
+ predictor = PredictorWrapperPtr(new PredictorWrapper());
+ predictor->load("./resources/model.pkl");
predictor->dump();
this->descriptor = {Size(64, 128), Size(16, 16), Size(8, 8), Size(8, 8), 9};
}
cv::Mat im2(PREFERRED_SIZE, p2->image_crop.type());
cv::resize(p1->image_crop, im1, im1.size());
cv::resize(p2->image_crop, im2, im2.size());
- cv::Mat result;
+
+ cv::Mat result = Mat::zeros(1,1, CV_64F);
+
cv::matchTemplate(im1, im2, result, CV_TM_CCOEFF_NORMED);
feature.push_back(result.at<double>(0, 0));
cv::matchTemplate(im1, im2, result, CV_TM_CCORR_NORMED);
feature.push_back(calc_iou_ratio(getRectInDetection(d1), getRectInDetection(d2)));
+// for (auto i : feature){
+// cout << i << "\t";
+// if (isnan(i)){
+// throw overflow_error("Nan in feature");
+// }
+// }
+// cout << endl;
+//
return feature;
}
ss = similarity(i, patch);
features.insert(features.end(), ss.begin(), ss.end());
}
- double prob = predictor->predict(4, features);
- return prob;
+ double prob = predictor->predict(tracker->patches.size() - 1, features); // TODO why is MaxPatch-1
+ if (prob > ProbThreshold)
+ return -log(prob);
+ else
+ return MaxCost;
}
static float calc_iou_ratio(const Detection& d1, const Detection& d2)
void MultiTracker::update(unsigned int total, const Detection* detections, const Mat& image)
{
+ // predict trackers, update trackers using kalman filter
+ for (auto t : trackers){
+ t->predict();
+ }
+
+ // match the trackers with the detections using linear sum assignment (hungarian)
int row = trackers.size();
int col = total;
Eigen::MatrixXi cost_matrix = Eigen::MatrixXi::Zero(row, col);
Eigen::VectorXi tracker_inds, bb_inds;
linear_sum_assignment(cost_matrix, tracker_inds, bb_inds);
- // handle unmatched trackers
- //vector<TrackerPtr> unmatched_trackers;
- for (int i = 0; i < row; i++){
+ set<TrackerPtr> unmatched_trackers;
+ set<int> unmatch_bbs_indices;
+
+ for(unsigned int i = 0; i < trackers.size(); i++){
if (!(tracker_inds.array() == i).any()){
- trackers[i]->updateState(image);
+ unmatched_trackers.insert(trackers[i]);
}
}
-
- // handle unmatched detections
- vector<int> unmatched_detection;
- for(int j = 0; j < col; j++){
+ for (unsigned int j = 0; j < total; j++){
if (!(bb_inds.array() == j).any()){
- unmatched_detection.push_back(j);
+ unmatch_bbs_indices.insert(j);
+ }
+ }
+
+ // handle matched trackers
+ for (unsigned int i = 0; i < tracker_inds.size(); i++){
+ for (int j = 0; j < bb_inds.size(); j++){
+ int rr = tracker_inds(i);
+ int cc = bb_inds(j);
+ TrackerPtr tracker = trackers[rr];
+ const Detection& detect = detections[cc];
+ if (cost_matrix(rr, cc) < MaxCost){
+ tracker->correct(image, detect);
+ tracker->addPatch(createPatch(image, detect));
+ } else {
+ unmatched_trackers.insert(tracker); // failed trackers
+ unmatch_bbs_indices.insert(cc); // filed detection
+ }
}
}
- // create new trackers for new detections
- for (auto i : unmatched_detection){
- TrackerPtr t (new Tracker(image));
- this->trackers.push_back(t);
+
+ // handle unmatched trackers
+ for (auto t : unmatched_trackers){
+ t->updateState(image);
+ }
+
+ // handle unmatched detections - Create new trackers
+ vector<Person> inPersons;
+ for (auto i : unmatch_bbs_indices){
+ TrackerPtr new_tracker (new Tracker(image, detections[i]));
+ new_tracker->addPatch(createPatch(image, detections[i]));
+ addTracker(new_tracker);
+ Person test; // TODO
+ inPersons.push_back(test);
+ }
+
+ // callback and notify engine - persons in
+ if (inPersons.size() > 0){
+ if (auto e = engine.lock()){
+ e->onPersonsIn(inPersons);
+ }
}
- Detection dd;
+ // Delete lost trackers
+ vector<Person> outPersons;
+ for (auto it = trackers.begin(); it < trackers.end(); it++){
+ if ((*it)->status == TrackerStatus::Delete){
+ Person test; // TODO
+ outPersons.push_back(test);
+ trackers.erase(it);
+ }
+ }
- PatchPtr pp = createPatch(image, dd);
+ // callback and notify engine - persons out
+ if (outPersons.size() > 0){
+ if (auto e = engine.lock()){
+ e->onPersonsOut(outPersons);
+ }
+ }
}
static cv::Mat image_crop(const cv::Mat& image, const Detection& bb)
patch->features = std::make_pair(feature_hog_double, hist);
return patch;
}
+
+void MultiTracker::addTracker(TrackerPtr t)
+{
+ trackers.insert(trackers.begin(), t);
+ if(trackers.size() > MaxTrackers){
+ LOG_ERROR(TAG, "trackers reaches the maximum " + to_string(MaxTrackers));
+ trackers.pop_back();
+ }
+}
projects / trackerpp.git / blobdiff