Multi-hypothesis 3D human pose estimation metrics favor miscalibrated distributions

TL;DR

This paper highlights the miscalibration issues in multi-hypothesis 3D human pose estimation metrics and introduces a new model, cGNF, that provides well-calibrated pose distributions while maintaining competitive accuracy.

Contribution

The paper identifies the cause of miscalibration in existing multi-hypothesis methods and proposes cGNF, a model that estimates both conditional and marginal densities for better calibration.

Findings

cGNF achieves well-calibrated pose distributions.

cGNF outperforms previous methods on occluded joints.

cGNF maintains competitive minMPJPE scores.

Abstract

Due to depth ambiguities and occlusions, lifting 2D poses to 3D is a highly ill-posed problem. Well-calibrated distributions of possible poses can make these ambiguities explicit and preserve the resulting uncertainty for downstream tasks. This study shows that previous attempts, which account for these ambiguities via multiple hypotheses generation, produce miscalibrated distributions. We identify that miscalibration can be attributed to the use of sample-based metrics such as minMPJPE. In a series of simulations, we show that minimizing minMPJPE, as commonly done, should converge to the correct mean prediction. However, it fails to correctly capture the uncertainty, thus resulting in a miscalibrated distribution. To mitigate this problem, we propose an accurate and well-calibrated model called Conditional Graph Normalizing Flow (cGNFs). Our model is structured such that a single cGNF can estimate both conditional and marginal densities within the same model - effectively solving a zero-shot density estimation problem. We evaluate cGNF on the Human~3.6M dataset and show that cGNF provides a well-calibrated distribution estimate while being close to state-of-the-art in terms of overall minMPJPE. Furthermore, cGNF outperforms previous methods on occluded joints while it remains well-calibrated.