Matlab implementation of the DCF-CSR tracker from the paper published in the proceedings of Conference on Computer Vision and Pattern Recognition (CVPR) 2017 and later in International Journal of Computer Vision (IJCV).
Journal publication:
Alan Lukežič, Tomáš Vojíř, Luka Čehovin, Jiří Matas and Matej Kristan. ''Discriminative Correlation Filter Tracker with Channel and Spatial Reliability.'' International Journal of Computer Vision (IJCV), 2018.
Paper
BibTex citation:
@Article{Lukezic_IJCV2018,
author={Luke{\v{z}}i{\v{c}}, Alan and Voj{'i}{\v{r}}, Tom{'a}{\v{s}} and {\v{C}}ehovin Zajc, Luka and Matas, Ji{\v{r}}{'i} and Kristan, Matej},
title={Discriminative Correlation Filter Tracker with Channel and Spatial Reliability},
journal={International Journal of Computer Vision},
year={2018},
}
Raw results: VOT15 VOT16 OTB100
Conference publication:
Alan Lukežič, Tomáš Vojíř, Luka Čehovin, Jiří Matas and Matej Kristan. ''Discriminative Correlation Filter with Channel and Spatial Reliability.'' In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017.
Paper
BibTex citation:
@InProceedings{Lukezic_CVPR_2017,
Title = {Discriminative Correlation Filter with Channel and Spatial Reliability},
Author = {Luke{\v{z}}i{\v{c}}, Alan and Voj{'i}{\v{r}}, Tom{'a}{\v{s}} and {\v{C}}ehovin Zajc, Luka and Matas, Ji{\v{r}}{'i} and Kristan, Matej},
Booktitle = {CVPR},
Year = {2017}
}
Alan Lukežič, e-mail: [email protected]
Tomáš Vojíř, e-mail: [email protected]
Luka Čehovin Zajc, e-mail: [email protected]
- Clone git repository:
$ git clone https://github.com/alanlukezic/csr-dcf.git - Compile mex files running compile.m command
Set opencv_include and opencv_libpath to the correct OpenCV paths - Use demo_csr.m script for the visualization of the tracker
Set tracker_path variable to the directory where your source code is and base_path to the directory where you have stored the VOT sequences. - Use csr_wrapper.m script for VOT integration
The C++ version of the CSR-DCF tracker is now available in OpenCV contrib repository (tracking module, CSRT tracker)
Short-term tracking is an open and challenging problem for which discriminative correlation filters (DCF) have shown excellent performance. We introduce the channel and spatial reliability concepts to DCF tracking and provide a novel learning algorithm for its efficient and seamless integration in the filter update and the tracking process. The spatial reliability map adjusts the filter support to the part of the object suitable for tracking. This both allows to enlarge the search region and improves tracking of non-rectangular objects. Reliability scores reflect channel-wise quality of the learned filters and are used as feature weighting coefficients in localization. Experimentally, with only two simple standard features, HoGs and Colornames, the novel CSR-DCF method -- DCF with Channel and Spatial Reliability -- achieves state-of-the-art results on VOT 2016, VOT 2015 and OTB100. The CSR-DCF runs in real-time on a CPU.
Tracking results on the VOT-2015 benchmark:
Tracking results on the VOT-2016 benchmark:
Click here to see demo video on YouTube.