This is a C++ implementation of the tracking algorithm described in the paper:
Struck: Structured Output Tracking with Kernels
Sam Hare, Amir Saffari, Philip H. S. Torr
International Conference on Computer Vision (ICCV), 2011
Tracking can be performed on video sequences, or live input from a webcam.
Copyright (C) 2011 Sam Hare, Oxford Brookes University, Oxford, UK
- OpenCV: http://opencv.org/
- Eigen: http://eigen.tuxfamily.org/
This code has been tested using OpenCV v2.4.12 and Eigen v3.2.6
CMake is used for cross-platform compilation. For example on Unix-based systems run:
> mkdir build
> cd build
> cmake ..
> make
Note: Make sure you compile the code in Release, as a Debug build will result in significantly slower performance
After compilation, from the top level of the repository run:
> build/bin/struck [config-file-path]
If no path is given the application will attempt to use ./config.txt.
Please see config.txt for configuration options.
Sequences are assumed to be of the format of those available from: http://vision.ucsd.edu/~bbabenko/project_miltrack.html
This code is released under the GPLv3 license for non-commercial use only. For other types of license please contact me.
This code uses the OpenCV graphing utilities provided by Shervin Emami: http://www.shervinemami.info/graphs.html