{"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/4d-spatio-temporal-convnets-minkowski","title":"4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks","arxiv_id":"1904.08755","date":"2019-04-18","proceeding":"CVPR 2019 6","authors":["Christopher Choy","JunYoung Gwak","Silvio Savarese"],"abstract":"In many robotics and VR/AR applications, 3D-videos are readily-available sources of input (a continuous sequence of depth images, or LIDAR scans). However, those 3D-videos are processed frame-by-frame either through 2D convnets or 3D perception algorithms. In this work, we propose 4-dimensional convolutional neural networks for spatio-temporal perception that can directly process such 3D-videos using high-dimensional convolutions. For this, we adopt sparse tensors and propose the generalized sparse convolution that encompasses all discrete convolutions. To implement the generalized sparse convolution, we create an open-source auto-differentiation library for sparse tensors that provides extensive functions for high-dimensional convolutional neural networks. We create 4D spatio-temporal convolutional neural networks using the library and validate them on various 3D semantic segmentation benchmarks and proposed 4D datasets for 3D-video perception. To overcome challenges in the 4D space, we propose the hybrid kernel, a special case of the generalized sparse convolution, and the trilateral-stationary conditional random field that enforces spatio-temporal consistency in the 7D space-time-chroma space. Experimentally, we show that convolutional neural networks with only generalized 3D sparse convolutions can outperform 2D or 2D-3D hybrid methods by a large margin. Also, we show that on 3D-videos, 4D spatio-temporal convolutional neural networks are robust to noise, outperform 3D convolutional neural networks and are faster than the 3D counterpart in some cases.","url_abs":"https://arxiv.org/abs/1904.08755v4","url_pdf":"https://arxiv.org/pdf/1904.08755v4.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":"4d-spatio-temporal-convnets-minkowski","repo_url":"https://github.com/StanfordVL/MinkowskiEngine","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"4d-spatio-temporal-convnets-minkowski","repo_url":"https://github.com/NVIDIA/MinkowskiEngine","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"4d-spatio-temporal-convnets-minkowski","repo_url":"https://github.com/buildingnet/buildingnet_dataset","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"4d-spatio-temporal-convnets-minkowski","repo_url":"https://github.com/dkoh0207/lartpc_minkowski","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"4d-spatio-temporal-convnets-minkowski","repo_url":"https://github.com/ldkong1205/Robo3D","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"4d-spatio-temporal-convnets-minkowski","repo_url":"https://github.com/mit-han-lab/spvnas","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"4d-spatio-temporal-convnets-minkowski","repo_url":"https://github.com/shwoo93/minkowskiengine","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"4d-spatio-temporal-convnets-minkowski","repo_url":"https://github.com/Pointcept/Pointcept","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"3d-semantic-segmentation","task_name":"3D Semantic Segmentation"},{"task_slug":"4d-spatio-temporal-semantic-segmentation","task_name":"4D Spatio Temporal Semantic Segmentation"},{"task_slug":"robust-3d-semantic-segmentation","task_name":"Robust 3D Semantic Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"sparse-convolutions","method_name":"Sparse Convolutions"},{"method_slug":"submanifold-convolutions","method_name":"Submanifold Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/3d-semantic-segmentation-on-stpls3d","task":"3D Semantic Segmentation","dataset":"STPLS3D","model":"MinkowskiNet","rank_in_archive_order":3,"of":6,"metrics":{"mIOU":"51.3"},"uses_additional_data":false},{"leaderboard":"/sota/3d-semantic-segmentation-on-scannet-1","task":"3D Semantic Segmentation","dataset":"ScanNet++","model":"SpUNet (MinkowskiNet)","rank_in_archive_order":7,"of":8,"metrics":{"Top-1 IoU":"0.456","Top-3 IoU":"0.683"},"uses_additional_data":false},{"leaderboard":"/sota/3d-semantic-segmentation-on-scannet200","task":"3D Semantic Segmentation","dataset":"ScanNet200","model":"MinkUNet","rank_in_archive_order":16,"of":16,"metrics":{"test mIoU":"25.3","val mIoU":"25.0"},"uses_additional_data":false},{"leaderboard":"/sota/3d-semantic-segmentation-on-scribblekitti","task":"3D Semantic Segmentation","dataset":"ScribbleKITTI","model":"MinkowskiNet","rank_in_archive_order":4,"of":6,"metrics":{"mIoU":"55.0"},"uses_additional_data":false},{"leaderboard":"/sota/3d-semantic-segmentation-on-wildscenes","task":"3D Semantic Segmentation","dataset":"WildScenes","model":"MinkUNet","rank_in_archive_order":3,"of":4,"metrics":{"mIoU":"36.53","mIoU (Env DA)":"30.78","mIoU (Temporal DA)":"27.20"},"uses_additional_data":false},{"leaderboard":"/sota/robust-3d-semantic-segmentation-on","task":"Robust 3D Semantic Segmentation","dataset":"SemanticKITTI-C","model":"MinkUNet-18","rank_in_archive_order":3,"of":22,"metrics":{"mean Corruption Error (mCE)":"100.00%"},"uses_additional_data":false},{"leaderboard":"/sota/robust-3d-semantic-segmentation-on","task":"Robust 3D Semantic Segmentation","dataset":"SemanticKITTI-C","model":"MinkUNet-34","rank_in_archive_order":5,"of":22,"metrics":{"mean Corruption Error (mCE)":"100.61%"},"uses_additional_data":false},{"leaderboard":"/sota/robust-3d-semantic-segmentation-on-wod-c","task":"Robust 3D Semantic Segmentation","dataset":"WOD-C","model":"MinkUNet-34","rank_in_archive_order":1,"of":5,"metrics":{"mean Corruption Error (mCE)":"96.21%"},"uses_additional_data":false},{"leaderboard":"/sota/robust-3d-semantic-segmentation-on-wod-c","task":"Robust 3D Semantic Segmentation","dataset":"WOD-C","model":"MinkUNet-18","rank_in_archive_order":3,"of":5,"metrics":{"mean Corruption Error (mCE)":"100.00%"},"uses_additional_data":false},{"leaderboard":"/sota/robust-3d-semantic-segmentation-on-nuscenes-c","task":"Robust 3D Semantic Segmentation","dataset":"nuScenes-C","model":"MinkUNet-34","rank_in_archive_order":2,"of":12,"metrics":{"mean Corruption Error (mCE)":"96.37%"},"uses_additional_data":false},{"leaderboard":"/sota/robust-3d-semantic-segmentation-on-nuscenes-c","task":"Robust 3D Semantic Segmentation","dataset":"nuScenes-C","model":"MinkUNet-18","rank_in_archive_order":5,"of":12,"metrics":{"mean Corruption Error (mCE)":"100.00%"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-s3dis","task":"Semantic Segmentation","dataset":"S3DIS","model":"MinkowskiNet","rank_in_archive_order":37,"of":54,"metrics":{"Mean IoU":"65.4","Number of params":"37.9M","Params (M)":"37.9"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-s3dis-area5","task":"Semantic Segmentation","dataset":"S3DIS Area5","model":"MinkowskiNet","rank_in_archive_order":45,"of":61,"metrics":{"Number of params":"37.9M","mAcc":"71.7","mIoU":"65.4"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-scannet","task":"Semantic Segmentation","dataset":"ScanNet","model":"MinkowskiNet","rank_in_archive_order":27,"of":45,"metrics":{"test mIoU":"73.4","val mIoU":"72.2"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1904.08755","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1904.08755"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ldkong1205/Robo3D","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/NVIDIA/MinkowskiEngine","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/shwoo93/minkowskiengine","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/buildingnet/buildingnet_dataset","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/dkoh0207/lartpc_minkowski","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/StanfordVL/MinkowskiEngine","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Pointcept/Pointcept","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/mit-han-lab/spvnas","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran_fixture":1,"unverified":1},"by_repo_kind":{"listed":{"samples":2,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":2,"samples":[{"code_sha256_prefix":"ba78a3f6ba016b4f","entry":"normalize_color","repo":"NVIDIA/MinkowskiEngine","repo_kind":"listed","path":"examples/indoor.py","file_url":"https://github.com/NVIDIA/MinkowskiEngine/blob/HEAD/examples/indoor.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"ba78a3f6ba016b4f"}},{"code_sha256_prefix":"460808fbce44a8d8","entry":"load_file","repo":"NVIDIA/MinkowskiEngine","repo_kind":"listed","path":"examples/indoor.py","file_url":"https://github.com/NVIDIA/MinkowskiEngine/blob/HEAD/examples/indoor.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"460808fbce44a8d8"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}