- Decision: conferenceOral-iclr2013-workshop
- Abstract: In this work, we consider the problem of detecting robotic grasps in an RGB-D view of a scene containing objects. We present a two-step cascaded structure, where we have two deep networks, with the top detections from the first one re-evaluated by the second one. The first deep network has fewer features, is therefore faster to run and makes more mistakes. The second network has more features and therefore gives better detections. Unlike previous works that need to design these features manually, deep learning gives us flexibility in designing such multi-step cascaded detectors.