Precise occlusion-aware and feature-level reconstruction for occluded person re-identification

Xiujun Shu, Hanjun Li, Wei Wen, Ruizhi Qiao, Nannan Li, Weijian Ruan, Hanjing Su, Bo Wang, Shouzhi Chen, Jun Zhou

Abstract: Highlights•A novel Precise Occlusion-aware and Feature-level Reconstruction (POFR) network is proposed, integrating an occlusion-driven contrastive learning (OCL) module and an occlusion-guided feature reconstruction (OFR) module for occluded person re-identification.•In OCL, an occlusion generator creates realistic occlusions and precise masks for OFR, while contrastive learning improves the model’s ability to learn occlusion-robust features from diverse occluded images.•In OFR, a lightweight occlusion predictor uses a precise mask to perceive occlusion information, aiding in the recovery of occluded features and reducing information asymmetry.