Object Drift Verification Network Based on Multi-Scale Feature Fusion and Dual-Template in Long-Term Tracking
Yu Chen, Zhiqiang Hou, Jiaxin Zhao, Sugang Ma, Wangsheng Yu, Yunchen Wang · 2024
To address the key challenge of verifying whether the tracking results have drifted in long-term tracking, we propose an object drift verification network based on multi-scale feature fusion and dual-template, using static and dynamic templates for joint verification.During the feature extraction stage, a multi-scale feature fusion module is introduced to adapt to changes in the object's scale.Additionally, A template update strategy is devised to obtain high-quality dynamic templates for effective object drift verification.This network doesn't require manual threshold setting and can be used as a plug-and-play module combined with a short-term visual tracking algorithm and a global re-detection module for long-term tracking.We validated the effectiveness of the proposed network using DiMP50 as the base tracker.Extensive experiments on the LaSOT, UAV20L, VOT2018-LT, and VOT2020-LT datasets demonstrate significant improvements in long-term tracking performance.Specifically, on the UAV20L dataset, success rate and precision improved by 8.3% and 11.3%, respectively, while the VOT2020-LT dataset saw a 5.3% increase in the F-score, compared to DiMP50.Moreover, the proposed network achieves a verification speed of 220 FPS with minimal impact on overall tracking speed.