APSIPA Transactions on Signal and Information Processing > Vol 11 > Issue 1

UHP-SOT++: An Unsupervised Lightweight Single Object Tracker

Zhiruo Zhou, University of Southern California, USA, zhiruozh@usc.edu , Hongyu Fu, University of Southern California, USA, Suya You, DEVCOM Army Research Laboratory, USA, C.-C. Jay Kuo, University of Southern California, USA
Suggested Citation
Zhiruo Zhou, Hongyu Fu, Suya You and C.-C. Jay Kuo (2022), "UHP-SOT++: An Unsupervised Lightweight Single Object Tracker", APSIPA Transactions on Signal and Information Processing: Vol. 11: No. 1, e27. http://dx.doi.org/10.1561/116.00000008

Publication Date: 01 Sep 2022
© 2022 Z. Zhou, H. Fu, S. You and C.-C. J. Kuo
Object trackingonline trackingsingle object trackingunsupervised tracking


Open Access

This is published under the terms of CC BY-NC.

Downloaded: 796 times

In this article:
Related Work 
Proposed UHP-SOT++ Method 
Exemplary Sequences and Qualitative Analysis 
Conclusion and Future Work 


An enhanced version of UHP-SOT called UHP-SOT++ is proposed for unsupervised, lightweight and high-performance single object tracking in this work. Both UHP-SOT and UHP-SOT++ exploit the discriminative-correlation-filters-based (DCF-based) tracker as their baseline and incorporate two new ingredients: (1) background motion modeling and (2) object box trajectory modeling. Their difference lies in the fusion strategy of proposals from three models (i.e., DCF, background motion and object box trajectory models). An improved fusion strategy is adopted by UHP-SOT++ for robust tracking performance against large-scale tracking datasets. Extensive evaluation of state-of-the-art supervised/unsupervised deep and unsupervised lightweight trackers is conducted on four SOT benchmark datasets – OTB2015, TC128, UAV123 and LaSOT. UHP-SOT++ achieves outstanding tracking performance with a small model size and low computational complexity (i.e., operating at a rate of 20 FPS on an i5 CPU even without code optimization). UHP-SOT++ offers an ideal solution in real-time object tracking on resource-limited platforms. Finally, we compare the pros and cons of supervised deep trackers and unsupervised lightweight trackers and provide a new perspective to their performance gap.