Temporal Context Attention-Based Tracker for UAV Tracking
Yuxiao Yin, Yungang Liu, Yongchao Man, Fengzhong Li, Yuan Wang · 2023
Inspired by the framework of TCTrack, a temporal context attention-based tracker is proposed for Unmanned Aerial Vehicle(UAV). The instruction of attention mechanism enables trackers to pay attention to more effective feature knowledge in order to compensate the problem that shallow network feature extraction of UAV tracker is limited, therefore the attention module of multi-feature information fusion based on temporal context is proposed to produce more distinguishable feature representation. In addition, considering the relationship between context similarities, we propose the lightweight temporal transformer(LT-Trans) to increase the execution speed of the model. Finally, the evaluation experiments of four air tracking benchmarks show our tracker compared with currently popular trackers makes more excellent performance.