Multiple-Human Parsing in the Wild

Human parsing is attracting increasing research attention. In this work, weaim to push the frontier of human parsing by introducing the problem ofmulti-human parsing in the wild. Existing works on human parsing mainly tacklesingle-person scenarios, which deviates from real-world applications wheremultiple persons are present simultaneously with interaction and occlusion. Toaddress the multi-human parsing problem, we introduce a new multi-human parsing(MHP) dataset and a novel multi-human parsing model named MH-Parser. The MHPdataset contains multiple persons captured in real-world scenes withpixel-level fine-grained semantic annotations in an instance-aware setting. TheMH-Parser generates global parsing maps and person instance maskssimultaneously in a bottom-up fashion with the help of a new Graph-GAN model.We envision that the MHP dataset will serve as a valuable data resource todevelop new multi-human parsing models, and the MH-Parser offers a strongbaseline to drive future research for multi-human parsing in the wild.