Siamese network for object tracking with multi-granularity appearance representations. (October 2021)

• Record Type: Journal Article
• Authors: Zhang, Zhuoyi; Zhang, Yifeng; Cheng, Xu; Lu, Guojun
• Is Part Of: Pattern recognition. Volume 118(2021)
• Publication Date: 2021-10
• Subjects: Object tracking -- Siamese network -- Appearance adaption; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2021.108003
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: Enhance semantic feature representations for targets in tracking by a improved Siamese network and a convolutional block attention module. Capture multi-granularity appearance features of objects at pixel, local and global level accurately. Combine two models dynamically with the Hedge algorithm in a selective traverse fashion efficiently to track objects. Design an updatable tracker which performs better that most state-of-the-art trackers and operates at speeds that exceed the real-time requirement. Abstract: A reliable tracker has the ability to adapt to change of objects over time, and is robust and accurate. We build such a tracker by extracting semantic features using robust Siamese networks and multi-granularity color features. It incorporates a semantic model that can capture high quality semantic features and an appearance model that can describe object at pixel, local and global levels effectively. Furthermore, we propose a novel selective traverse algorithm to allocate weights to semantic models and appearance models dynamically for better tracking performance. During tracking, our tracker updates appearance representations for objects based on the recent tracking results. The proposed tracker operates at speeds that exceed the real-time requirement, and outperforms nearly all other state-of-the-art trackers on OTB-2013/2015 and VOT-2016/2017 benchmarks.