Efficient 3D Pose Estimation and 3D Model Retrieval

Retrieving 3D models for objects in 2D images is an increasingly important problem, driven by the recent emergence of large databases of 3D models such as ShapeNet. However, this task is challenging for two main reasons: (1) 2D images and 3D models have considerably different representations and characteristics, making it hard to compare them. (2) The appearance of objects can significantly vary with the pose, but it is in general unknown and, thus, multiple poses have to be considered, which is very inefficient. To overcome these problems, we propose first to predict the object's pose and then to use the estimated pose as a prior to retrieve 3D models from a database. In the following, we give a short summary of the approach in Sec.~\ref{sec:3D-pose} and a sketch of results in Sec.~\ref{sec:results}. For more details, we refer to Grabner et.al. "3D Pose Estimation and 3D Model Retrieval for Objects in the Wild".