We present a method for the learning and detection of multiple rigid texture-less 3Dobjects intended to operate at frame rate speeds for video input. The method is gearedfor fast and scalable learning and detection by combining tractable extraction of edgeletconstellations with library lookup based on rotation- and scale-invariant descriptors. Theapproach learns object views in real-time, and is generative - enabling more objects tobe learnt without the need for re-training. During testing, a random sample of edgeletconstellations is tested for the presence of known objects. We perform testing of singleand multi-object detection on a 30 objects dataset showing detections of any of themwithin milliseconds from the object’s visibility. The results show the scalability of theapproach and its framerate performance.
展开▼
