{"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/capnet-continuous-approximation-projection","title":"CAPNet: Continuous Approximation Projection For 3D Point Cloud Reconstruction Using 2D Supervision","arxiv_id":"1811.11731","date":"2018-11-28","proceeding":null,"authors":["Navaneet K L","Priyanka Mandikal","Mayank Agarwal","R. Venkatesh Babu"],"abstract":"Knowledge of 3D properties of objects is a necessity in order to build\neffective computer vision systems. However, lack of large scale 3D datasets can\nbe a major constraint for data-driven approaches in learning such properties.\nWe consider the task of single image 3D point cloud reconstruction, and aim to\nutilize multiple foreground masks as our supervisory data to alleviate the need\nfor large scale 3D datasets. A novel differentiable projection module, called\n'CAPNet', is introduced to obtain such 2D masks from a predicted 3D point\ncloud. The key idea is to model the projections as a continuous approximation\nof the points in the point cloud. To overcome the challenges of sparse\nprojection maps, we propose a loss formulation termed 'affinity loss' to\ngenerate outlier-free reconstructions. We significantly outperform the existing\nprojection based approaches on a large-scale synthetic dataset. We show the\nutility and generalizability of such a 2D supervised approach through\nexperiments on a real-world dataset, where lack of 3D data can be a serious\nconcern. To further enhance the reconstructions, we also propose a test stage\noptimization procedure to obtain reconstructions that display high\ncorrespondence with the observed input image.","url_abs":"http://arxiv.org/abs/1811.11731v1","url_pdf":"http://arxiv.org/pdf/1811.11731v1.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":"capnet-continuous-approximation-projection","repo_url":"https://github.com/val-iisc/capnet","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"3d-point-cloud-reconstruction","task_name":"3D Point Cloud Reconstruction"},{"task_slug":"point-cloud-reconstruction","task_name":"Point cloud reconstruction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1811.11731","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}