Improving 2D Human Pose Estimation in Rare Camera Views with Synthetic Data

TL;DR

This paper introduces RePoGen, a synthetic data generator that improves 2D human pose estimation in rare camera views, especially top- and bottom-views, by augmenting existing datasets without degrading performance on common views.

Contribution

The paper presents RePoGen, a novel SMPL-based synthetic data generator that enhances pose estimation in rare camera views and demonstrates its effectiveness through experiments and dataset creation.

Findings

Adding RePoGen data improves top- and bottom-view pose estimation.

Anatomical plausibility is not essential for effective synthetic data.

RePoGen outperforms previous methods in rare view scenarios.

Abstract

Methods and datasets for human pose estimation focus predominantly on side- and front-view scenarios. We overcome the limitation by leveraging synthetic data and introduce RePoGen (RarE POses GENerator), an SMPL-based method for generating synthetic humans with comprehensive control over pose and view. Experiments on top-view datasets and a new dataset of real images with diverse poses show that adding the RePoGen data to the COCO dataset outperforms previous approaches to top- and bottom-view pose estimation without harming performance on common views. An ablation study shows that anatomical plausibility, a property prior research focused on, is not a prerequisite for effective performance. The introduced dataset and the corresponding code are available on https://mirapurkrabek.github.io/RePoGen-paper/ .