Regression trackers directly learn a mapping from regularly dense samples of target objects to soft labels, which are usually generated by a Gaussian function, to estimate target positions. Due to the potential for fast-tracking and easy implementation, regression trackers have recently received increasing attention. However, state-of-the-art deep regression trackers do not perform as well as discriminative correlation filters (DCFs) trackers. We identify the main bottleneck of training regression networks as extreme foreground-background data imbalance. To balance training data, we propose a novel shrinkage loss to penalize the importance of easy training data. Additionally, we apply residual connections to fuse multiple convolutional layers as well as their output response maps. Without bells and whistles, the proposed deep regression tracking method performs favorably against state-of-the-art trackers, especially in comparison with DCFs trackers, on five benchmark datasets including OTB-2013, OTB-2015, Temple-128, UAV-123 and VOT-2016.
|Title of host publication||Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings|
|Editors||Vittorio Ferrari, Cristian Sminchisescu, Yair Weiss, Martial Hebert|
|Number of pages||18|
|Publication status||Published - 2018|
|Event||15th European Conference on Computer Vision, ECCV 2018 - Munich, Germany|
Duration: 2018 Sep 8 → 2018 Sep 14
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Other||15th European Conference on Computer Vision, ECCV 2018|
|Period||18/9/8 → 18/9/14|
Bibliographical noteFunding Information:
Acknowledgments. This work is supported in part by the National Key Research and Development Program of China (2016YFB1001003), NSFC (61527804, 61521062, U1611461, 61502301, and 61671298), the 111 Program (B07022), and STCSM (17511105401 and 18DZ2270700). C. Ma and I. Reid acknowledge the support of the Australian Research Council through the Centre of Excellence for Robotic Vision (CE140100016) and Laureate Fellowship (FL130100102). B. Ni is supported by China’s Thousand Youth Talents Plan. M.-H. Yang is supported by NSF CAREER (1149783).
© 2018, Springer Nature Switzerland AG.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)