One-Shot Transfer of Affordance Regions? AffCorrs!
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.
PDF AbstractCode
Tasks
Object
One-Shot Instance Segmentation
One-Shot Part Segmentation of Contain Affordance - Inter Class
One-Shot Part Segmentation of Contain Affordance - Intra Class
One-Shot Part Segmentation of Cut Affordance - Inter Class
One-Shot Part Segmentation of Cut Affordance - Intra Class
One-Shot Part Segmentation of Grasp Affordance - Inter Class
One-Shot Part Segmentation of Grasp Affordance - Intra Class
One-Shot Part Segmentation of Pound Affordance - Inter Class
One-Shot Part Segmentation of Pound Affordance - Intra Class
One-Shot Part Segmentation of Scoop Affordance - Inter Class
One-Shot Part Segmentation of Scoop Affordance - Intra Class
One-Shot Part Segmentation of Support Affordance - Inter Class
One-Shot Part Segmentation of Support Affordance - Intra Class
One-Shot Part Segmentation of Wrap-Grasp Affordance - Inter Class
One-Shot Part Segmentation of Wrap-Grasp Affordance - Intra Class
One-Shot Segmentation
UMD-i Affrodance Dataset
