Back to the Future: Learning Shape Models from 3D CAD Data

Recognizing 3D objects from arbitrary view points is one of the fundamental problems in computer vision. A major challenge lies the transition between the 3D geometry of objects and 2D that can be robustly matched to natural images. Most thus rely on 2D natural images either as the sole source of data for building an implicit 3D representation, or by 3D models with natural image features. this paper, we go back to the ideas from the early days of computer, by using 3D object models as the only source of information for a multi-view object class detector. In particular, we use models for learning 2D shape that can be robustly matched to 2D images. Our experiments confirm the validity of our approach, outperforms current state-of-the-art techniques on a multi-view data set.