Multiple-Human Parsing in the Wild

TL;DR

This paper introduces the multi-human parsing problem in real-world scenes, presents a new dataset with detailed annotations, and proposes a novel model that jointly generates global parsing maps and individual person masks.

Contribution

It provides the first multi-human parsing dataset with pixel-level annotations and a new bottom-up model using Graph-GAN for simultaneous global and instance-aware parsing.

Findings

The MHP dataset enables research on multi-human parsing in complex scenes.

MH-Parser achieves competitive performance on the new dataset.

The approach effectively handles occlusion and interaction among multiple persons.

Abstract

Human parsing is attracting increasing research attention. In this work, we aim to push the frontier of human parsing by introducing the problem of multi-human parsing in the wild. Existing works on human parsing mainly tackle single-person scenarios, which deviates from real-world applications where multiple persons are present simultaneously with interaction and occlusion. To address 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 MHP dataset contains multiple persons captured in real-world scenes with pixel-level fine-grained semantic annotations in an instance-aware setting. The MH-Parser generates global parsing maps and person instance masks simultaneously 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 to develop new multi-human parsing models, and the MH-Parser offers a strong baseline to drive future research for multi-human parsing in the wild.