Tracking in tracking: An efficient method to solve the tracking distortion

Jinzhen Yao, Zhixing Wang, Jianlin Zhang, Qintao Hu, Chuanming Tang, Qiliang Bao, Zhenming Peng

Abstract: Highlights•Current Siamese trackers find it hard to discriminate similar distractors.•The improvement in trackers’ performance heavily relies on the enormous model complexity and parameters.•We first proposed a tracking-in-tracking framework to solve them.•By utilizing our framework, it realizes stable tracking with high robustness.