One-Shot Transfer of Affordance Regions? AffCorrs!

Denis Hadjivelichkov, Sicelukwanda Zwane, Lourdes Agapito, Marc Peter Deisenroth, Dimitrios Kanoulas

Student First Author: yes

Keywords: One-shot, Affordance, Correspondence

TL;DR: One-shot transfer of affordance masks to unseen objects, both belonging to the same class and other classes.

Abstract: In this work, we tackle one-shot visual search of object parts. Given a single reference image of an object with annotated affordance regions, we segment semantically corresponding parts within a target scene. We propose AffCorrs, an unsupervised model that combines the properties of pre-trained DINO-ViT's image descriptors and cyclic correspondences. We use AffCorrs to find corresponding affordances both for intra- and inter-class one-shot part segmentation. This task is more difficult than supervised alternatives, but enables future work such as learning affordances via imitation and assisted teleoperation.

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2209.07147/code)

Website: https://sites.google.com/view/affcorrs

Code: https://github.com/RPL-CS-UCL/UCL-AffCorrs