Rethinking Keypoint Representations: Modeling Keypoints and Poses as Objects for Multi-Person Human Pose Estimation

TL;DR

KAPAO introduces a novel object-based approach for multi-person human pose estimation, replacing heatmap regression with a more efficient, accurate, and unified detection framework that models keypoints and poses as objects.

Contribution

The paper presents KAPAO, a new single-stage dense detection method that models keypoints and poses as objects, improving speed and accuracy over traditional heatmap-based methods.

Findings

KAPAO outperforms previous methods in accuracy and speed.

The object-based approach reduces computation and quantization errors.

KAPAO performs well without test-time augmentation.

Abstract

In keypoint estimation tasks such as human pose estimation, heatmap-based regression is the dominant approach despite possessing notable drawbacks: heatmaps intrinsically suffer from quantization error and require excessive computation to generate and post-process. Motivated to find a more efficient solution, we propose to model individual keypoints and sets of spatially related keypoints (i.e., poses) as objects within a dense single-stage anchor-based detection framework. Hence, we call our method KAPAO (pronounced "Ka-Pow"), for Keypoints And Poses As Objects. KAPAO is applied to the problem of single-stage multi-person human pose estimation by simultaneously detecting human pose and keypoint objects and fusing the detections to exploit the strengths of both object representations. In experiments, we observe that KAPAO is faster and more accurate than previous methods, which suffer greatly from heatmap post-processing. The accuracy-speed trade-off is especially favourable in the practical setting when not using test-time augmentation. Source code: https://github.com/wmcnally/kapao.