Deep intra-image contrastive learning for weakly supervised one-step person search

Jiabei Wang, Yanwei Pang, Jiale Cao, Hanqing Sun, Zhuang Shao, Xuelong Li

Abstract: Highlights•We propose a novel weakly supervised one-step person search framework via deep intra-image contrastive learning (DICL).•We design two novel intra-image contrast modules, including spatial-invariant contrast (SIC) and occlusion-invariant contrast (OIC).•Without using the identity annotations, DICL fully mines the intra-image information to deal with spatial-level and occlusion-level variance.