tyb197/

RoboAfford-Eval

The RoboAfford-Eval Benchmark is designed to evaluate object and spatial affordance understanding abilities for enhancing robot manipulation. This comprehensive benchmark consists of:

114 questions for object affordance recognition tasks

124 questions for object affordance prediction tasks

100 questions for spatial affordance localization tasks (adapted from Where2Place)

The benchmark utilizes images from the Where2Place dataset and features… See the full description on the dataset page: https://huggingface.co/datasets/tyb197/RoboAfford-Eval.