AMST: Object tracking based on collaborative framework with adaptive multi-strategy

Rui Xu, Si Chen, Yan Yan, Da-Han Wang, Shunzhi Zhu

Abstract: Highlights•An adaptive multi-strategy collaborative framework is proposed for tracking.•A variable action set module updates multiple strategies dramatically.•A template update and selection module manages reliable templates dynamically.•An online reliability evaluation module employs different evaluation functions.•An offline learning Siamese tracker is developed for multi-template matching.