Retail Objects Affordance Dataset, underlying the MSc thesis: Object Affordance Detection for Mobile Manipulation in Retail Environments

Synthetically generated affordance dataset consisting of 3237 images with pixel-level annotations for either graspable or pushable affordance classes. Further, 204 annotated real images of similar scenarios are included.Images contain retail store related objects lying of supermarket floors, together with a semantic masks differentiating the object from the background.

本数据集为合成生成的可供性（affordance）数据集，包含3237张带有像素级标注的图像，标注类别涵盖可抓取可供性与可推动可供性两类。此外，数据集还收录了204张对应相似场景的标注真实图像。图像内容为放置在超市地面的零售相关物体，同时附带用于区分物体与背景的语义掩码（semantic masks）。