Learning Motion-Perceive Siamese network for robust visual object tracking

Ze Kang, Tianyang Xu, Xuefeng Zhu, Xiaojun Wu

Abstract: Highlights•A novel Motion-Perceive Siamese network.•A motion branch is proposed to predict the motion state of the target.•An adaptive fusion module is proposed to fuse classification maps from two branches.•The proposed method achieves competitive performance on several tracking datasets.