SA6D: Self-Adaptive Few-Shot 6D Pose Estimator for Novel and Occluded Objects
It addresses the challenge of generalizable and scalable pose estimation for novel objects under occlusion in robotics, representing a novel method for a known bottleneck.
The paper tackles the problem of 6D pose estimation for novel and occluded objects in robotic manipulation by proposing SA6D, a few-shot approach that uses self-adaptive segmentation and point cloud modeling from cluttered images, and demonstrates it outperforms existing methods in cluttered scenes with fewer reference images.
To enable meaningful robotic manipulation of objects in the real-world, 6D pose estimation is one of the critical aspects. Most existing approaches have difficulties to extend predictions to scenarios where novel object instances are continuously introduced, especially with heavy occlusions. In this work, we propose a few-shot pose estimation (FSPE) approach called SA6D, which uses a self-adaptive segmentation module to identify the novel target object and construct a point cloud model of the target object using only a small number of cluttered reference images. Unlike existing methods, SA6D does not require object-centric reference images or any additional object information, making it a more generalizable and scalable solution across categories. We evaluate SA6D on real-world tabletop object datasets and demonstrate that SA6D outperforms existing FSPE methods, particularly in cluttered scenes with occlusions, while requiring fewer reference images.