关键词：外观；视觉；线索；3D
摘 要：In this paper, we investigate the prediction of visual grasp affordances from 2D measurements. Appearancebased estimation of grasp affordances is desirable when 3-D scans are unreliable due to clutter or material properties.We develop a general framework for estimating grasp affordances from 2-D sources, including local texture-like measures as well as object-category measures that capture previously learned grasp strategies. Local approaches to estimating grasp positions have been shown to be effective in real-world scenarios, but are unable to impart objectlevel biases and can be prone to false positives. We describe how global cues can be used to compute continuous pose estimates and corresponding grasp point locations, using a max-margin optimization for category-level continuous pose regression.