{"url":"/sota/3d-instance-segmentation-on-scannetv2","task":{"name":"3D Instance Segmentation","url":"/task/3d-instance-segmentation-1","note":null},"dataset":{"name":"ScanNet(v2)","url":"/dataset/scannet"},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"Image: [OccuSeg](https://arxiv.org/pdf/2003.06537v3.pdf)","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["mAP @ 50","mAP@25","mAP","mRec"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"mAP @ 50":"higher","mAP@25":"higher","mAP":"higher","mRec":null}},"counts":{"rows":32,"rows_with_code":23,"rows_with_paper_page":32,"rows_dated":32,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"Relation3D","metrics":{"mAP":"62.2","mAP @ 50":"81.6","mAP@25":"90.1"},"uses_additional_data":false,"paper_date":"2025-01-01","paper":"/paper/relation3d-enhancing-relation-modeling-for","paper_url":"http://openaccess.thecvf.com//content/CVPR2025/html/Lu_Relation3D__Enhancing_Relation_Modeling_for_Point_Cloud_Instance_Segmentation_CVPR_2025_paper.html","paper_title":"Relation3D : Enhancing Relation Modeling for Point Cloud Instance Segmentation","code":"https://github.com/Howard-coder191/Relation3D","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"Spherical Mask","metrics":{"mAP":"61.6","mAP @ 50":"81.2","mAP@25":"87.5"},"uses_additional_data":false,"paper_date":"2023-12-18","paper":"/paper/spherical-mask-coarse-to-fine-3d-point-cloud","paper_url":"https://arxiv.org/abs/2312.11269v2","paper_title":"Spherical Mask: Coarse-to-Fine 3D Point Cloud Instance Segmentation with Spherical Representation","code":"https://github.com/yunshin/SphericalMask","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":3,"n_samples":4,"n_pointer_only_licence":0}},{"rank_in_archive_order":3,"model":"BFL","metrics":{"mAP":"60.6","mAP @ 50":"81.0","mAP@25":"88.2"},"uses_additional_data":false,"paper_date":"2025-02-06","paper":"/paper/beyond-the-final-layer-hierarchical-query","paper_url":"https://arxiv.org/abs/2502.04139v1","paper_title":"Beyond the Final Layer: Hierarchical Query Fusion Transformer with Agent-Interpolation Initialization for 3D Instance Segmentation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"OneFromer3D","metrics":{"mAP":"56.6","mAP @ 50":"80.1","mAP@25":"89.6"},"uses_additional_data":false,"paper_date":"2023-11-24","paper":"/paper/oneformer3d-one-transformer-for-unified-point","paper_url":"https://arxiv.org/abs/2311.14405v1","paper_title":"OneFormer3D: One Transformer for Unified Point Cloud Segmentation","code":"https://github.com/oneformer3d/oneformer3d","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"MSTA3D","metrics":{"mAP":"56.9","mAP @ 50":"79.5","mAP@25":"87.9","mRec":"74.1"},"uses_additional_data":false,"paper_date":"2024-11-04","paper":"/paper/msta3d-multi-scale-twin-attention-for-3d-1","paper_url":"https://arxiv.org/abs/2411.01781v3","paper_title":"MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":6,"model":"QueryFormer","metrics":{"mAP":"58.3","mAP @ 50":"78.7","mAP@25":"87.4"},"uses_additional_data":false,"paper_date":"2023-01-01","paper":"/paper/query-refinement-transformer-for-3d-instance","paper_url":"http://openaccess.thecvf.com//content/ICCV2023/html/Lu_Query_Refinement_Transformer_for_3D_Instance_Segmentation_ICCV_2023_paper.html","paper_title":"Query Refinement Transformer for 3D Instance Segmentation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"Mask3D","metrics":{"mAP":"55.2","mAP @ 50":"78.0","mAP@25":"87.0"},"uses_additional_data":false,"paper_date":"2022-10-06","paper":"/paper/mask3d-for-3d-semantic-instance-segmentation","paper_url":"https://arxiv.org/abs/2210.03105v2","paper_title":"Mask3D: Mask Transformer for 3D Semantic Instance Segmentation","code":"https://github.com/jonasschult/mask3d","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":8,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"SPFormer","metrics":{"mAP":"54.9","mAP @ 50":"77.0"},"uses_additional_data":false,"paper_date":"2022-11-28","paper":"/paper/superpoint-transformer-for-3d-scene-instance","paper_url":"https://arxiv.org/abs/2211.15766v1","paper_title":"Superpoint Transformer for 3D Scene Instance Segmentation","code":"https://github.com/sunjiahao1999/spformer","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"ISBNet","metrics":{"mAP":"55.9","mAP @ 50":"76.3","mAP@25":"84.5"},"uses_additional_data":false,"paper_date":"2023-03-01","paper":"/paper/isbnet-a-3d-point-cloud-instance-segmentation","paper_url":"https://arxiv.org/abs/2303.00246v2","paper_title":"ISBNet: a 3D Point Cloud Instance Segmentation Network with Instance-aware Sampling and Box-aware Dynamic Convolution","code":"https://github.com/VinAIResearch/ISBNet","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":17,"n_samples":19,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"SoftGroup","metrics":{"mAP":"50.4","mAP @ 50":"76.1","mAP@25":"86.5"},"uses_additional_data":false,"paper_date":"2022-03-03","paper":"/paper/softgroup-for-3d-instance-segmentation-on","paper_url":"https://arxiv.org/abs/2203.01509v1","paper_title":"SoftGroup for 3D Instance Segmentation on Point Clouds","code":"https://github.com/thangvubk/softgroup","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"TD3D","metrics":{"mAP":"48.9","mAP @ 50":"75.1","mAP@25":"87.5"},"uses_additional_data":false,"paper_date":"2023-02-06","paper":"/paper/top-down-beats-bottom-up-in-3d-instance","paper_url":"https://arxiv.org/abs/2302.02871v4","paper_title":"Top-Down Beats Bottom-Up in 3D Instance Segmentation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":12,"model":"PBNet","metrics":{"mAP":"57.3","mAP @ 50":"74.7","mAP@25":"82.5"},"uses_additional_data":false,"paper_date":"2022-07-22","paper":"/paper/divide-and-conquer-3d-point-cloud-instance","paper_url":"https://arxiv.org/abs/2207.11209v4","paper_title":"Divide and Conquer: 3D Point Cloud Instance Segmentation With Point-Wise Binarization","code":"https://github.com/weiguangzhao/PBNet","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":4,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":13,"model":"DKNet","metrics":{"mAP":"53.2","mAP @ 50":"71.8"},"uses_additional_data":false,"paper_date":"2022-07-15","paper":"/paper/3d-instances-as-1d-kernels","paper_url":"https://arxiv.org/abs/2207.07372v2","paper_title":"3D Instances as 1D Kernels","code":"https://github.com/w1zheng/dknet","n_code_links":1,"syntology":{"n_ran":6,"n_unverified":0,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"ODIN","metrics":{"mAP":"50.0","mAP @ 50":"71.0","mAP@25":"83.6"},"uses_additional_data":false,"paper_date":"2024-01-04","paper":"/paper/odin-a-single-model-for-2d-and-3d-perception","paper_url":"https://arxiv.org/abs/2401.02416v3","paper_title":"ODIN: A Single Model for 2D and 3D Segmentation","code":"https://github.com/ayushjain1144/odin","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":5,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":15,"model":"HAIS","metrics":{"mAP":"45.7","mAP @ 50":"69.9"},"uses_additional_data":false,"paper_date":"2021-08-05","paper":"/paper/hierarchical-aggregation-for-3d-instance","paper_url":"https://arxiv.org/abs/2108.02350v1","paper_title":"Hierarchical Aggregation for 3D Instance Segmentation","code":"https://github.com/hustvl/HAIS","n_code_links":1,"syntology":null},{"rank_in_archive_order":16,"model":"SSTNet","metrics":{"mAP":"50.6","mAP @ 50":"69.8"},"uses_additional_data":false,"paper_date":"2021-08-17","paper":"/paper/instance-segmentation-in-3d-scenes-using","paper_url":"https://arxiv.org/abs/2108.07478v1","paper_title":"Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks","code":"https://github.com/gorilla-lab-scut/sstnet","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":8,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":17,"model":"MaskGroup","metrics":{"mAP":"43.4","mAP @ 50":"66.4"},"uses_additional_data":false,"paper_date":"2022-03-28","paper":"/paper/maskgroup-hierarchical-point-grouping-and","paper_url":"https://arxiv.org/abs/2203.14662v1","paper_title":"MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":18,"model":"RPGN","metrics":{"mAP":"42.8","mAP @ 50":"64.2","mAP@25":"80.6"},"uses_additional_data":false,"paper_date":"2020-11-03","paper":"/paper/learning-regional-purity-for-instance","paper_url":"https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900055.pdf","paper_title":"Learning Regional Purity for Instance Segmentation on 3D Point Clouds","code":"https://github.com/dsc1126/RPGN","n_code_links":1,"syntology":null},{"rank_in_archive_order":19,"model":"GICN","metrics":{"mAP @ 50":"63.8"},"uses_additional_data":false,"paper_date":"2020-07-20","paper":"/paper/learning-gaussian-instance-segmentation-in","paper_url":"https://arxiv.org/abs/2007.09860v1","paper_title":"Learning Gaussian Instance Segmentation in Point Clouds","code":"https://github.com/LiuShihHung/GICN","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":20,"model":"PointGroup","metrics":{"mAP":"40.7","mAP @ 50":"63.6"},"uses_additional_data":false,"paper_date":"2020-04-03","paper":"/paper/pointgroup-dual-set-point-grouping-for-3d","paper_url":"https://arxiv.org/abs/2004.01658v1","paper_title":"PointGroup: Dual-Set Point Grouping for 3D Instance Segmentation","code":"https://github.com/Pointcept/Pointcept","n_code_links":4,"syntology":{"n_ran":0,"n_unverified":5,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":21,"model":"HIDA","metrics":{"mAP":"43.6","mAP @ 50":"63.5"},"uses_additional_data":false,"paper_date":"2021-07-07","paper":"/paper/hida-towards-holistic-indoor-understanding","paper_url":"https://arxiv.org/abs/2107.03180v1","paper_title":"HIDA: Towards Holistic Indoor Understanding for the Visually Impaired via Semantic Instance Segmentation with a Wearable Solid-State LiDAR Sensor","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":22,"model":"3D-MPA","metrics":{"mAP":"35.3","mAP @ 50":"59.1","mRec":"61.1"},"uses_additional_data":false,"paper_date":"2020-03-30","paper":"/paper/3d-mpa-multi-proposal-aggregation-for-3d","paper_url":"https://arxiv.org/abs/2003.13867v1","paper_title":"3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation","code":"https://github.com/francisengelmann/3D-MPA","n_code_links":1,"syntology":null},{"rank_in_archive_order":23,"model":"4DContrast","metrics":{"mAP @ 50":"57.6"},"uses_additional_data":false,"paper_date":"2021-12-06","paper":"/paper/4dcontrast-contrastive-learning-with-dynamic","paper_url":"https://arxiv.org/abs/2112.02990v2","paper_title":"4DContrast: Contrastive Learning with Dynamic Correspondences for 3D Scene Understanding","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":24,"model":"RWSeg","metrics":{"mAP":"34.8","mAP @ 50":"56.7","mAP@25":"73.9"},"uses_additional_data":false,"paper_date":"2022-08-10","paper":"/paper/rwseg-cross-graph-competing-random-walks-for","paper_url":"https://arxiv.org/abs/2208.05110v3","paper_title":"Collaborative Propagation on Multiple Instance Graphs for 3D Instance Segmentation with Single-point Supervision","code":"https://github.com/dsc1126/RWSeg","n_code_links":1,"syntology":null},{"rank_in_archive_order":25,"model":"3D-BoNet","metrics":{"mAP":"25.3","mAP @ 50":"48.8","mRec":"47.6"},"uses_additional_data":false,"paper_date":"2019-06-04","paper":"/paper/learning-object-bounding-boxes-for-3d","paper_url":"https://arxiv.org/abs/1906.01140v2","paper_title":"Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds","code":"https://github.com/Yang7879/3D-BoNet","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":2,"n_samples":4,"n_pointer_only_licence":0}},{"rank_in_archive_order":26,"model":"ResNet-Backbone","metrics":{"mAP @ 50":"45.9"},"uses_additional_data":false,"paper_date":"2019-02-12","paper":"/paper/masc-multi-scale-affinity-with-sparse","paper_url":"http://arxiv.org/abs/1902.04478v1","paper_title":"MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation","code":"https://github.com/art-programmer/MASC","n_code_links":1,"syntology":null},{"rank_in_archive_order":27,"model":"MASC","metrics":{"mAP":"25.4","mAP @ 50":"44.7"},"uses_additional_data":false,"paper_date":"2019-02-12","paper":"/paper/masc-multi-scale-affinity-with-sparse","paper_url":"http://arxiv.org/abs/1902.04478v1","paper_title":"MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation","code":"https://github.com/art-programmer/MASC","n_code_links":1,"syntology":null},{"rank_in_archive_order":28,"model":"3D-SIS","metrics":{"mAP @ 50":"38.2"},"uses_additional_data":false,"paper_date":"2018-12-17","paper":"/paper/3d-sis-3d-semantic-instance-segmentation-of","paper_url":"http://arxiv.org/abs/1812.07003v3","paper_title":"3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans","code":"https://github.com/Sekunde/3D-SIS","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":29,"model":"UNet-Backbone","metrics":{"mAP @ 50":"31.9"},"uses_additional_data":false,"paper_date":"2016-06-21","paper":"/paper/3d-u-net-learning-dense-volumetric","paper_url":"http://arxiv.org/abs/1606.06650v1","paper_title":"3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation","code":"https://github.com/wolny/pytorch-3dunet","n_code_links":27,"syntology":{"n_ran":53,"n_unverified":23,"n_samples":76,"n_pointer_only_licence":20}},{"rank_in_archive_order":30,"model":"MPNet","metrics":{"mAP @ 50":"31"},"uses_additional_data":false,"paper_date":"2020-01-06","paper":"/paper/learning-and-memorizing-representative","paper_url":"https://arxiv.org/abs/2001.01349v1","paper_title":"Learning and Memorizing Representative Prototypes for 3D Point Cloud Semantic and Instance Segmentation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":31,"model":"Searilized Point Mamba","metrics":{"mAP":"40.0","mAP@25":"76.4"},"uses_additional_data":false,"paper_date":"2024-07-17","paper":"/paper/serialized-point-mamba-a-serialized-point","paper_url":"https://arxiv.org/abs/2407.12319v1","paper_title":"Serialized Point Mamba: A Serialized Point Cloud Mamba Segmentation Model","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":32,"model":"MAFT","metrics":{"mAP":"59.6"},"uses_additional_data":false,"paper_date":"2023-09-04","paper":"/paper/mask-attention-free-transformer-for-3d","paper_url":"https://arxiv.org/abs/2309.01692v1","paper_title":"Mask-Attention-Free Transformer for 3D Instance Segmentation","code":"https://github.com/dvlab-research/mask-attention-free-transformer","n_code_links":1,"syntology":{"n_ran":4,"n_unverified":6,"n_samples":10,"n_pointer_only_licence":10}}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,264 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9581,"papers_checked":6264,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":3316},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":13,"rows_with_any_sample_ran":9,"distinct_papers_with_graph_line":13,"distinct_papers_with_any_sample_ran":9,"samples_over_distinct_papers":{"n_ran":75,"n_unverified":84,"n_samples":159,"n_pointer_only_licence":33,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":75,"n_unverified":84,"n_samples":159,"n_pointer_only_licence":33,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}