Project page of WSCF (Fast Learning of Spatially Regularized and Content Aware Correlation Filter for Visual Tracking), published in TIP 2020.
@article{han2020fast,
title={Fast Learning of Spatially Regularized and Content Aware Correlation Filter for Visual Tracking},
author={Han, Ruize and Feng, Wei and and Wang, Song},
year={2020},
journal={IEEE Transactions on Image Processing}
}
In this paper, we propose a new fast learning approach to content-aware spatial regularization, namely weighted sample based CF tracking (WSCF). In WSCF, specifically, we present a simple yet effective energy function that implicitly weighs different training samples by spatial deviations. With the energy function, the learning of correlation filters is composed of two subproblems with closed-form solution and can be efficiently solved in an alternate way. We further develop a content-aware updating strategy to dynamically refine the weight distribution to well adapt to the temporal variations of the target and background.
The experimental results on OTB-2013/2015: (We provide the raw results on OTB benchmark in 'results_WSCF_OTB-2015.zip'.)