Self-supervised Point Set Local Descriptors for Point Cloud Registration
This addresses the need for efficient point cloud registration in domains like robotics and computer vision by eliminating the requirement for labeled data, though it builds incrementally on prior work.
The paper tackles the problem of learning local descriptors for point cloud registration without manual annotation by proposing a self-supervised method that trains on a single unlabeled point cloud using self-rotation supervision. The results show that this method achieves equivalent or better performance than supervised models on various datasets.
In this work, we propose to learn local descriptors for point clouds in a self-supervised manner. In each iteration of the training, the input of the network is merely one unlabeled point cloud. On top of our previous work, that directly solves the transformation between two point sets in one step without correspondences, the proposed method is able to train from one point cloud, by supervising its self-rotation, that we randomly generate. The whole training requires no manual annotation. In several experiments we evaluate the performance of our method on various datasets and compare to other state of the art algorithms. The results show, that our self-supervised learned descriptor achieves equivalent or even better performance than the supervised learned model, while being easier to train and not requiring labeled data.