gengshan-y / headTracking

Code for detection-based head tracking.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

headTracking

C++ implementation for multi-target tracking algorithm from paper Multi-target tracking by learning local-to-global trajectory models.
Using Kalman filter and SVM classifier to generate tracking scores.

Note

  • imgSVM class trains/tests SVM classifier through images.
  • Tracker is a higher level class including implementation of Kalman Filters and call of imgSVM class.
  • cvLib implements methods for updating trackers and for other utilities specific for tracking task.
  • cmpLib implements methods of image feature extraction and comparison.
  • global.hpp/.cpp keeps intersection of global variables for both imgSVM class and tracking task.

Usage

  • ./headTracking /data/gengshan/vid/testMultiTarget.avi y

todo

  • use distance to leverage svm score...
  • divide updateTracker function
  • optimzie display setting... put tracking result in detection result later.

Log

  • params: negNum in TrackingObj class.

About

Code for detection-based head tracking.


Languages

Language:C++ 96.9%Language:Makefile 1.7%Language:Shell 1.4%