{"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/tangent-convolutions-for-dense-prediction-in","title":"Tangent Convolutions for Dense Prediction in 3D","arxiv_id":"1807.02443","date":"2018-07-06","proceeding":"CVPR 2018 6","authors":["Maxim Tatarchenko","Jaesik Park","Vladlen Koltun","Qian-Yi Zhou"],"abstract":"We present an approach to semantic scene analysis using deep convolutional\nnetworks. Our approach is based on tangent convolutions - a new construction\nfor convolutional networks on 3D data. In contrast to volumetric approaches,\nour method operates directly on surface geometry. Crucially, the construction\nis applicable to unstructured point clouds and other noisy real-world data. We\nshow that tangent convolutions can be evaluated efficiently on large-scale\npoint clouds with millions of points. Using tangent convolutions, we design a\ndeep fully-convolutional network for semantic segmentation of 3D point clouds,\nand apply it to challenging real-world datasets of indoor and outdoor 3D\nenvironments. Experimental results show that the presented approach outperforms\nother recent deep network constructions in detailed analysis of large 3D\nscenes.","url_abs":"http://arxiv.org/abs/1807.02443v1","url_pdf":"http://arxiv.org/pdf/1807.02443v1.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":"tangent-convolutions-for-dense-prediction-in","repo_url":"https://github.com/tatarchm/tangent_conv","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"3d-semantic-segmentation","task_name":"3D Semantic Segmentation"},{"task_slug":"prediction","task_name":"Prediction"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/3d-semantic-segmentation-on-semantickitti","task":"3D Semantic Segmentation","dataset":"SemanticKITTI","model":"TangentConv","rank_in_archive_order":36,"of":45,"metrics":{"test mIoU":"35.9%"},"uses_additional_data":false},{"leaderboard":"/sota/3d-semantic-segmentation-on-sensaturban","task":"3D Semantic Segmentation","dataset":"SensatUrban","model":"TangentConv","rank_in_archive_order":8,"of":8,"metrics":{"mIoU":"33.30"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-s3dis-area5","task":"Semantic Segmentation","dataset":"S3DIS Area5","model":"TangentConv","rank_in_archive_order":58,"of":61,"metrics":{"Number of params":"N/A","mAcc":"62.2"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-scannet","task":"Semantic Segmentation","dataset":"ScanNet","model":"Tangent Convolutions","rank_in_archive_order":43,"of":45,"metrics":{"test mIoU":"44.2"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1807.02443","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}