{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/non-rigid-point-set-registration-networks","title":"Non-Rigid Point Set Registration Networks","arxiv_id":"1904.01428","date":"2019-04-02","proceeding":null,"authors":["Lingjing Wang","Jianchun Chen","Xiang Li","Yi Fang"],"abstract":"Point set registration is defined as a process to determine the spatial\ntransformation from the source point set to the target one. Existing methods\noften iteratively search for the optimal geometric transformation to register a\ngiven pair of point sets, driven by minimizing a predefined alignment loss\nfunction. In contrast, the proposed point registration neural network (PR-Net)\nactively learns the registration pattern as a parametric function from a\ntraining dataset, consequently predict the desired geometric transformation to\nalign a pair of point sets. PR-Net can transfer the learned knowledge (i.e.\nregistration pattern) from registering training pairs to testing ones without\nadditional iterative optimization. Specifically, in this paper, we develop\nnovel techniques to learn shape descriptors from point sets that help formulate\na clear correlation between source and target point sets. With the defined\ncorrelation, PR-Net tends to predict the transformation so that the source and\ntarget point sets can be statistically aligned, which in turn leads to an\noptimal spatial geometric registration. PR-Net achieves robust and superior\nperformance for non-rigid registration of point sets, even in presence of\nGaussian noise, outliers, and missing points, but requires much less time for\nregistering large number of pairs. More importantly, for a new pair of point\nsets, PR-Net is able to directly predict the desired transformation using the\nlearned model without repetitive iterative optimization routine. Our code is\navailable at https://github.com/Lingjing324/PR-Net.","url_abs":"http://arxiv.org/abs/1904.01428v1","url_pdf":"http://arxiv.org/pdf/1904.01428v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"non-rigid-point-set-registration-networks","repo_url":"https://github.com/Lingjing324/PR-Net","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1904.01428","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}