Human Pose Regression by Combining Indirect Part Detection and Contextual Information

TL;DR

This paper introduces a fully differentiable end-to-end human pose regression method that leverages indirect heat map learning and contextual information, achieving state-of-the-art results on challenging datasets.

Contribution

It presents a novel Soft-argmax based regression framework that learns heat maps indirectly without extra ground truth, enabling seamless integration of contextual cues.

Findings

Achieved top performance on LSP and MPII datasets.

Outperformed existing regression methods in human pose estimation.

Comparable to detection-based approaches in accuracy.

Abstract

In this paper, we propose an end-to-end trainable regression approach for human pose estimation from still images. We use the proposed Soft-argmax function to convert feature maps directly to joint coordinates, resulting in a fully differentiable framework. Our method is able to learn heat maps representations indirectly, without additional steps of artificial ground truth generation. Consequently, contextual information can be included to the pose predictions in a seamless way. We evaluated our method on two very challenging datasets, the Leeds Sports Poses (LSP) and the MPII Human Pose datasets, reaching the best performance among all the existing regression methods and comparable results to the state-of-the-art detection based approaches.