{"ID":2854466,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.14354","arxiv_id":"2510.14354","title":"Leveraging Cycle-Consistent Anchor Points for Self-Supervised RGB-D Registration","abstract":"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 meth- ods 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 self- supervised registration methods on ScanNet and 3DMatch, even outperforming some older supervised methods. We also integrate our components into existing methods, showing their effectiveness.","short_abstract":"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 meth- ods rely on geometric and feature-based similarity, we take a different approach. We use cycle-cons...","url_abs":"https://arxiv.org/abs/2510.14354","url_pdf":"https://arxiv.org/pdf/2510.14354v1","authors":"[\"Siddharth Tourani\",\"Jayaram Reddy\",\"Sarvesh Thakur\",\"K Madhava Krishna\",\"Muhammad Haris Khan\",\"N Dinesh Reddy\"]","published":"2025-10-16T06:47:10Z","proceeding":"cs.CV","tasks":"[\"cs.CV\",\"cs.RO\"]","methods":"[]","has_code":false}