A Simple Method to Boost Human Pose Estimation Accuracy by Correcting the Joint Regressor for the Human3.6m Dataset

TL;DR

This paper introduces a method to improve human joint estimation accuracy by refining the joint regressor using pseudo-ground-truth SMPL poses, leading to better pose results on the Human3.6m dataset without retraining.

Contribution

The authors propose a novel approach to enhance the SMPL-to-joint regressor by generating pseudo-ground-truth SMPL poses, significantly improving pose estimation accuracy.

Findings

More accurate joint locations on Human3.6m

Improved pose estimation results without retraining

Open-source code and regressor released

Abstract

Many human pose estimation methods estimate Skinned Multi-Person Linear (SMPL) models and regress the human joints from these SMPL estimates. In this work, we show that the most widely used SMPL-to-joint linear layer (joint regressor) is inaccurate, which may mislead pose evaluation results. To achieve a more accurate joint regressor, we propose a method to create pseudo-ground-truth SMPL poses, which can then be used to train an improved regressor. Specifically, we optimize SMPL estimates coming from a state-of-the-art method so that its projection matches the silhouettes of humans in the scene, as well as the ground-truth 2D joint locations. While the quality of this pseudo-ground-truth is challenging to assess due to the lack of actual ground-truth SMPL, with the Human 3.6m dataset, we qualitatively show that our joint locations are more accurate and that our regressor leads to improved pose estimations results on the test set without any need for retraining. We release our code and joint regressor at https://github.com/ubc-vision/joint-regressor-refinement