AROS: Affordance Recognition with One-Shot Human Stances

TL;DR

AROS is a one-shot learning method that predicts human-scene affordances and generates articulated human poses in 3D environments using minimal examples, outperforming data-intensive baselines.

Contribution

It introduces a novel one-shot approach for affordance recognition that does not require re-training for new instances and effectively predicts human poses in unseen 3D scenes.

Findings

Outperforms data-intensive baselines by up to 80%

Works effectively with minimal examples of target poses

Validated on three public datasets with real environment scans

Abstract

We present AROS, a one-shot learning approach that uses an explicit representation of interactions between highly-articulated human poses and 3D scenes. The approach is one-shot as the method does not require re-training to add new affordance instances. Furthermore, only one or a small handful of examples of the target pose are needed to describe the interaction. Given a 3D mesh of a previously unseen scene, we can predict affordance locations that support the interactions and generate corresponding articulated 3D human bodies around them. We evaluate on three public datasets of scans of real environments with varied degrees of noise. Via rigorous statistical analysis of crowdsourced evaluations, results show that our one-shot approach outperforms data-intensive baselines by up to 80\%.