HuManiFlow: Ancestor-Conditioned Normalising Flows on SO(3) Manifolds for Human Pose and Shape Distribution Estimation

TL;DR

HuManiFlow introduces a novel probabilistic method for 3D human pose and shape estimation that balances accuracy, consistency, and diversity by modeling ancestor-conditioned distributions on SO(3) manifolds.

Contribution

It proposes a new approach using ancestor-conditioned normalising flows on SO(3) manifolds to improve diversity and consistency in 3D human pose estimation.

Findings

HuManiFlow achieves more accurate 3D pose distributions.

The method maintains high sample diversity.

It outperforms existing probabilistic models in consistency and diversity.

Abstract

Monocular 3D human pose and shape estimation is an ill-posed problem since multiple 3D solutions can explain a 2D image of a subject. Recent approaches predict a probability distribution over plausible 3D pose and shape parameters conditioned on the image. We show that these approaches exhibit a trade-off between three key properties: (i) accuracy - the likelihood of the ground-truth 3D solution under the predicted distribution, (ii) sample-input consistency - the extent to which 3D samples from the predicted distribution match the visible 2D image evidence, and (iii) sample diversity - the range of plausible 3D solutions modelled by the predicted distribution. Our method, HuManiFlow, predicts simultaneously accurate, consistent and diverse distributions. We use the human kinematic tree to factorise full body pose into ancestor-conditioned per-body-part pose distributions in an autoregressive manner. Per-body-part distributions are implemented using normalising flows that respect the manifold structure of SO(3), the Lie group of per-body-part poses. We show that ill-posed, but ubiquitous, 3D point estimate losses reduce sample diversity, and employ only probabilistic training losses. Code is available at: https://github.com/akashsengupta1997/HuManiFlow.