BoIR: Box-Supervised Instance Representation for Multi-Person Pose Estimation

TL;DR

BoIR introduces a bounding box-level instance representation learning method for multi-person pose estimation that improves feature disentanglement and detection in crowded scenes without extra inference cost.

Contribution

It proposes a novel instance embedding loss and multi-task learning framework that enhances multi-person pose estimation under crowded conditions.

Findings

Outperforms state-of-the-art on COCO val and test-dev datasets.

Achieves significant improvements on CrowdPose and OCHuman datasets.

Effective in disentangling instances in crowded scenes.

Abstract

Single-stage multi-person human pose estimation (MPPE) methods have shown great performance improvements, but existing methods fail to disentangle features by individual instances under crowded scenes. In this paper, we propose a bounding box-level instance representation learning called BoIR, which simultaneously solves instance detection, instance disentanglement, and instance-keypoint association problems. Our new instance embedding loss provides a learning signal on the entire area of the image with bounding box annotations, achieving globally consistent and disentangled instance representation. Our method exploits multi-task learning of bottom-up keypoint estimation, bounding box regression, and contrastive instance embedding learning, without additional computational cost during inference. BoIR is effective for crowded scenes, outperforming state-of-the-art on COCO val (0.8 AP), COCO test-dev (0.5 AP), CrowdPose (4.9 AP), and OCHuman (3.5 AP). Code will be available at https://github.com/uyoung-jeong/BoIR