Learning diverse fine-grained features for thermal infrared tracking

Chao Yang, Qiao Liu, Gaojun Li, Honghu Pan, Zhenyu He

Abstract: Highlights•Infrared fine-grained features can improve the performance of target tracking.•The network can extract infrared fine-grained features without labels.•The proposed DFG loss can improve the diversity of fine-grained features.•Infrared targets require strong features because of the lack of details.