Explaining the Ambiguity of Object Detection and 6D Pose From Visual Data
This addresses uncertainty in computer vision tasks like robotics and AR, but is incremental as it builds on existing methods to handle ambiguity.
The paper tackles the inherent ambiguity in 3D object detection and pose estimation from single images by predicting multiple pose and class outcomes to estimate pose distributions, resulting in higher accuracy in pose estimation.
3D object detection and pose estimation from a single image are two inherently ambiguous problems. Oftentimes, objects appear similar from different viewpoints due to shape symmetries, occlusion and repetitive textures. This ambiguity in both detection and pose estimation means that an object instance can be perfectly described by several different poses and even classes. In this work we propose to explicitly deal with this uncertainty. For each object instance we predict multiple pose and class outcomes to estimate the specific pose distribution generated by symmetries and repetitive textures. The distribution collapses to a single outcome when the visual appearance uniquely identifies just one valid pose. We show the benefits of our approach which provides not only a better explanation for pose ambiguity, but also a higher accuracy in terms of pose estimation.