This tutorial aims to provide the necessary background for understanding different indoor problems, their difficulties, the different sensors and data sources that one can use, how to exploit them, as well as how to formulate the problems such that efficient learning and inference is possible. In particular, we will review models for 3D object detection in both, monocular and RGB-D imagery, semantic segmentation (class and instance level), room layout estimation, as well as depth and normal estimation from a single image. The tutorial will discuss different possible parameterizations, sensors, the use of generative vs discriminative models, learned vs hand-crafted features, use of extra-information (e.g., CAD models, furniture catalogues), possible inference techniques, and possible learning algorithms. This will allow the audience to get a bigger picture of what is happening in the field, and what is potentially missing.
This tutorial is held on Sunday, June 7 at 2pm - 6pm in room 202. There will a coffee break at 3:45pm - 4:15pm (30 mins).
|Introduction, basic geometry, monocular 3D detection||Sanja||[pdf]  (99Mb)|
|Room layout estimation (with different parametrizations), holistic indoor models||Raquel||[pdf]  (18Mb)|
|Reconstruction and localization, semantics in RGB-D||Sanja||[pdf]  (64Mb)|