Point-level feature learning based on vision transformer for occluded person re-identification

Hua Gao, Chenchen Hu, Guang Han, Jiafa Mao, Wei Huang, Qiu Guan

Abstract: Highlights•A point-level person feature extractor is proposed for occluded person images.•Designing a part-based Transformer branch to improve person re-identification.•Combining point-level, part-level, and global features to present a person.•Experimental results show that our method is effective on person re-identification.