Preface

摘要

Welcome to the Proceedings for the 2nd Workshop on Recovering 6D Object Pose, held in conjunction with the European Conference on Computer Vision on October 9th, 2016. The proposed workshop is the second attempt at gathering researchers working on 6 DoF object pose estimation, 3D object detection and registration problems. The concerned topic is re-emerging in the field of computer vision and robotics after the development of RGB-D sensors and high GPU computing that have opened the door to a whole new range of technologies and applications which require detecting and estimating object poses in 3D environments for a variety of scenarios, e.g. picking and placing for logistics. The workshop (a) gathers people from both industry and academia who work on this topic, (b) help share knowledge and up-to-date advances, and (c) discuss benchmarks, baselines and evaluation metrics in the community. A majority of existing data sets and methods are limited on a certain degree of background clutters and foreground occlusions, and scenarios of a single target instance per image. The community needs to make one step forward and, thus our goal is to evaluate more challenging and realistic scenarios, where multiple target objects appear interacting and occluding each other or cases of articulated objects. We also aim at comparing new and established methods based on RGB and/or RGBD for object pose estimation on varied datasets covering multiple scenarios and settings.