{"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/fpnn-field-probing-neural-networks-for-3d","title":"FPNN: Field Probing Neural Networks for 3D Data","arxiv_id":"1605.06240","date":"2016-05-20","proceeding":"NeurIPS 2016 12","authors":["Yangyan Li","Soeren Pirk","Hao Su","Charles R. Qi","Leonidas J. Guibas"],"abstract":"Building discriminative representations for 3D data has been an important\ntask in computer graphics and computer vision research. Convolutional Neural\nNetworks (CNNs) have shown to operate on 2D images with great success for a\nvariety of tasks. Lifting convolution operators to 3D (3DCNNs) seems like a\nplausible and promising next step. Unfortunately, the computational complexity\nof 3D CNNs grows cubically with respect to voxel resolution. Moreover, since\nmost 3D geometry representations are boundary based, occupied regions do not\nincrease proportionately with the size of the discretization, resulting in\nwasted computation. In this work, we represent 3D spaces as volumetric fields,\nand propose a novel design that employs field probing filters to efficiently\nextract features from them. Each field probing filter is a set of probing\npoints --- sensors that perceive the space. Our learning algorithm optimizes\nnot only the weights associated with the probing points, but also their\nlocations, which deforms the shape of the probing filters and adaptively\ndistributes them in 3D space. The optimized probing points sense the 3D space\n\"intelligently\", rather than operating blindly over the entire domain. We show\nthat field probing is significantly more efficient than 3DCNNs, while providing\nstate-of-the-art performance, on classification tasks for 3D object recognition\nbenchmark datasets.","url_abs":"http://arxiv.org/abs/1605.06240v3","url_pdf":"http://arxiv.org/pdf/1605.06240v3.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":"fpnn-field-probing-neural-networks-for-3d","repo_url":"https://github.com/yangyanli/FPNN","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"fpnn-field-probing-neural-networks-for-3d","repo_url":"https://github.com/MindSpore-scientific-2/code-3/tree/main/ssFPN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":"3d-object-recognition","task_name":"3D Object Recognition"},{"task_slug":"3d-geometry","task_name":"3D geometry"},{"task_slug":"object-recognition","task_name":"Object Recognition"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/3d-object-recognition-on-modelnet40","task":"3D Object Recognition","dataset":"ModelNet40","model":"FPNN (4-FCs + NF)","rank_in_archive_order":5,"of":6,"metrics":{"Accuracy":"88.4%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1605.06240","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}