Towards Good Practices for Multi-Person Pose Estimation

TL;DR

This paper explores best practices for multi-person pose estimation by refining existing models and empirically evaluating their effects, achieving state-of-the-art results on COCO datasets.

Contribution

It introduces a series of refinements to MSPN and PoseFix networks and provides an empirical evaluation of their impact on pose estimation performance.

Findings

Achieved 78.7 AP on COCO test-dev dataset.

Achieved 76.3 AP on COCO test-challenge dataset.

Demonstrated the effectiveness of model refinements through ablation studies.

Abstract

Multi-Person Pose Estimation is an interesting yet challenging task in computer vision. In this paper, we conduct a series of refinements with the MSPN and PoseFix Networks, and empirically evaluate their impact on the final model performance through ablation studies. By taking all the refinements, we achieve 78.7 on the COCO test-dev dataset and 76.3 on the COCO test-challenge dataset.