{"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/fully-convolutional-point-networks-for-large","title":"Fully-Convolutional Point Networks for Large-Scale Point Clouds","arxiv_id":"1808.06840","date":"2018-08-21","proceeding":"ECCV 2018 9","authors":["Dario Rethage","Johanna Wald","Jürgen Sturm","Nassir Navab","Federico Tombari"],"abstract":"This work proposes a general-purpose, fully-convolutional network\narchitecture for efficiently processing large-scale 3D data. One striking\ncharacteristic of our approach is its ability to process unorganized 3D\nrepresentations such as point clouds as input, then transforming them\ninternally to ordered structures to be processed via 3D convolutions. In\ncontrast to conventional approaches that maintain either unorganized or\norganized representations, from input to output, our approach has the advantage\nof operating on memory efficient input data representations while at the same\ntime exploiting the natural structure of convolutional operations to avoid the\nredundant computing and storing of spatial information in the network. The\nnetwork eliminates the need to pre- or post process the raw sensor data. This,\ntogether with the fully-convolutional nature of the network, makes it an\nend-to-end method able to process point clouds of huge spaces or even entire\nrooms with up to 200k points at once. Another advantage is that our network can\nproduce either an ordered output or map predictions directly onto the input\ncloud, thus making it suitable as a general-purpose point cloud descriptor\napplicable to many 3D tasks. We demonstrate our network's ability to\neffectively learn both low-level features as well as complex compositional\nrelationships by evaluating it on benchmark datasets for semantic voxel\nsegmentation, semantic part segmentation and 3D scene captioning.","url_abs":"http://arxiv.org/abs/1808.06840v1","url_pdf":"http://arxiv.org/pdf/1808.06840v1.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":"fully-convolutional-point-networks-for-large","repo_url":"https://github.com/drethage/fully-convolutional-point-network","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/semantic-segmentation-on-scannet","task":"Semantic Segmentation","dataset":"ScanNet","model":"FCPN","rank_in_archive_order":41,"of":45,"metrics":{"test mIoU":"44.7"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1808.06840","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}