Leveraging Cycle-Consistent Anchor Points for Self-Supervised RGB-D Registration

Siddharth Tourani∗§1    Jayaram Reddy†2    Sarvesh Thakur†2    K Madhava Krishna†2    Muhammad Haris Khan§3    N Dinesh Reddy‡4   

1 Computer Vision and Learning Lab, University of Heidelberg    2 Robotics Research Center, IIIT Hyderabad, India    3 MBZUAI    4 Amazon   


With the rise in consumer depth cameras, a wealth of unlabeled RGB-D data has become available. This prompts the question of how to utilize this data for geometric reasoning of scenes. While many RGB-D registration methods rely on geometric and feature-based similarity, we take a different approach. We use cycle-consistent keypoints as salient points to enforce spatial coherence constraints during matching, improving correspondence accuracy. Additionally, we introduce a novel pose block that combines a GRU recurrent unit with transformation synchronization, blending historical and multi-view data. Our approach surpasses previous selfsupervised registration methods on ScanNet and 3DMatch, even outperforming some older supervised methods. We also integrate our components into existing methods, showing their effectiveness.