SiamVGG: Visual Tracking using Deeper Siamese Networks

• DOI: 10.48550/arxiv.1902.02804
• Publication Date: 2019-01-01
• Authors: Yuhong Li; Xiaofan Zhang
• Keywords: BitTorrent tracker; Discriminative model; Tracking (education); Frame rate

Recently, we have seen a rapid development of Deep Neural Network (DNN) based visual tracking solutions. Some trackers combine the DNN-based solutions with Discriminative Correlation Filters (DCF) to extract semantic features and successfully deliver state-of-the-art accuracy. However, these are highly compute-intensive, which require long processing time, resulting unsecured real-time performance. To both high accuracy reliable performance, propose novel tracker called SiamVGG\footnote{https://github.com/leeyeehoo/SiamVGG}. It combines Convolutional (CNN) backbone cross-correlation operator, takes advantage from exemplary images for more accurate object tracking. The architecture SiamVGG is customized VGG-16 parameters shared by desired input video frames. We demonstrate proposed on OTB-2013/50/100 VOT 2015/2016/2017 datasets while maintaining decent performance 50 FPS running GTX 1080Ti. Our design can achieve 2% higher Expected Average Overlap (EAO) compared ECO C-COT in VOT2017 Challenge.