ABOS: an attention-based one-stage framework for person search

Abstract Person search is of great significance to public safety research, such as crime surveillance, video surveillance and security. a method locating identifying the queried person from complete set images. The main cause false recall missed detection in presence occlusion In order improve accuracy when be occluded, this paper proposes an attention-based one-stage framework for (ABOS) using anchor-free model baseline. uses channel attention module express different forms take full advantage spatial highlight target region occluded pedestrians. These modules integrate deep shallow features guide network pay visible area extract semantic information Experimental results on CUHK-SYSU PRW datasets show that proposed based mechanism has better performance than existing methods, achieving 93.7 $$\%$$ % mAP dataset 46.4 dataset, respectively.