3D Semantic Label Results - ScanNet Benchmark

The 3D semantic labeling task involves predicting a semantic labeling of a 3D scan mesh.

Evaluation and metrics

Our evaluation ranks all methods according to the PASCAL VOC intersection-over-union metric (IoU). IoU = TP/(TP+FP+FN), where TP, FP, and FN are the numbers of true positive, false positive, and false negative pixels, respectively. Predicted labels are evaluated per-vertex over the respective 3D scan mesh; for 3D approaches that operate on other representations like grids or points, the predicted labels should be mapped onto the mesh vertices (e.g., one such example for grid to mesh vertices is provided in the evaluation helpers).

This table lists the benchmark results for the 3D semantic label scenario.

Method	avg iou	bathtub	bed	bookshelf	cabinet	chair	counter	curtain	desk	door	floor	otherfurniture	picture	refrigerator	shower curtain	sink	sofa	table	toilet	wall	window

PonderV2	0.785 2	0.978 1	0.800 26	0.833 22	0.788 4	0.853 16	0.545 16	0.910 5	0.713 1	0.705 4	0.979 1	0.596 7	0.390 1	0.769 11	0.832 41	0.821 4	0.792 29	0.730 1	0.975 1	0.897 4	0.785 4
Haoyi Zhu, Honghui Yang, Xiaoyang Wu, Di Huang, Sha Zhang, Xianglong He, Tong He, Hengshuang Zhao, Chunhua Shen, Yu Qiao, Wanli Ouyang: PonderV2: Pave the Way for 3D Foundataion Model with A Universal Pre-training Paradigm.
OccuSeg+Semantic	0.764 9	0.758 58	0.796 30	0.839 18	0.746 24	0.907 1	0.562 11	0.850 24	0.680 16	0.672 14	0.978 4	0.610 3	0.335 17	0.777 6	0.819 44	0.847 1	0.830 1	0.691 14	0.972 2	0.885 8	0.727 21

TTT-KD	0.773 5	0.646 90	0.818 14	0.809 34	0.774 9	0.878 3	0.581 2	0.943 1	0.687 12	0.704 5	0.978 4	0.607 5	0.336 15	0.775 8	0.912 6	0.838 3	0.823 2	0.694 12	0.967 3	0.899 2	0.794 3
Lisa Weijler, Muhammad Jehanzeb Mirza, Leon Sick, Can Ekkazan, Pedro Hermosilla: TTT-KD: Test-Time Training for 3D Semantic Segmentation through Knowledge Distillation from Foundation Models.
PTv3 ScanNet	0.794 1	0.941 3	0.813 18	0.851 7	0.782 6	0.890 2	0.597 1	0.916 2	0.696 8	0.713 3	0.979 1	0.635 1	0.384 2	0.793 2	0.907 8	0.821 4	0.790 30	0.696 11	0.967 3	0.903 1	0.805 1
Xiaoyang Wu, Li Jiang, Peng-Shuai Wang, Zhijian Liu, Xihui Liu, Yu Qiao, Wanli Ouyang, Tong He, Hengshuang Zhao: Point Transformer V3: Simpler, Faster, Stronger. CVPR 2024 (Oral)
PNE	0.755 14	0.786 42	0.835 4	0.834 21	0.758 15	0.849 21	0.570 8	0.836 31	0.648 26	0.668 16	0.978 4	0.581 16	0.367 6	0.683 34	0.856 29	0.804 6	0.801 19	0.678 18	0.961 5	0.889 5	0.716 28
P. Hermosilla: Point Neighborhood Embeddings.
O3DSeg	0.668 55	0.822 33	0.771 45	0.496 104	0.651 51	0.833 38	0.541 18	0.761 54	0.555 68	0.611 36	0.966 13	0.489 48	0.370 5	0.388 98	0.580 80	0.776 13	0.751 55	0.570 64	0.956 6	0.817 53	0.646 50

PPT-SpUNet-Joint	0.766 7	0.932 5	0.794 32	0.829 24	0.751 22	0.854 14	0.540 20	0.903 7	0.630 33	0.672 14	0.963 14	0.565 21	0.357 8	0.788 3	0.900 12	0.737 25	0.802 15	0.685 16	0.950 7	0.887 6	0.780 5
Xiaoyang Wu, Zhuotao Tian, Xin Wen, Bohao Peng, Xihui Liu, Kaicheng Yu, Hengshuang Zhao: Towards Large-scale 3D Representation Learning with Multi-dataset Point Prompt Training. CVPR 2024
IPCA	0.731 31	0.890 14	0.837 3	0.864 2	0.726 31	0.873 5	0.530 25	0.824 36	0.489 86	0.647 20	0.978 4	0.609 4	0.336 15	0.624 49	0.733 57	0.758 19	0.776 36	0.570 64	0.949 8	0.877 13	0.728 19

PointConvFormer	0.749 18	0.793 40	0.790 35	0.807 36	0.750 23	0.856 13	0.524 26	0.881 13	0.588 52	0.642 26	0.977 8	0.591 10	0.274 46	0.781 4	0.929 3	0.804 6	0.796 23	0.642 33	0.947 9	0.885 8	0.715 29
Wenxuan Wu, Qi Shan, Li Fuxin: PointConvFormer: Revenge of the Point-based Convolution.
One-Thing-One-Click	0.693 42	0.743 65	0.794 32	0.655 84	0.684 41	0.822 49	0.497 41	0.719 67	0.622 35	0.617 34	0.977 8	0.447 69	0.339 14	0.750 26	0.664 73	0.703 43	0.790 30	0.596 54	0.946 10	0.855 31	0.647 49
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
PointTransformerV2	0.752 16	0.742 66	0.809 21	0.872 1	0.758 15	0.860 11	0.552 13	0.891 12	0.610 40	0.687 6	0.960 17	0.559 24	0.304 29	0.766 14	0.926 4	0.767 16	0.797 22	0.644 32	0.942 11	0.876 16	0.722 25
Xiaoyang Wu, Yixing Lao, Li Jiang, Xihui Liu, Hengshuang Zhao: Point Transformer V2: Grouped Vector Attention and Partition-based Pooling. NeurIPS 2022
Retro-FPN	0.744 24	0.842 27	0.800 26	0.767 54	0.740 26	0.836 35	0.541 18	0.914 3	0.672 18	0.626 31	0.958 20	0.552 27	0.272 48	0.777 6	0.886 19	0.696 45	0.801 19	0.674 24	0.941 12	0.858 27	0.717 26
Peng Xiang, Xin Wen, Yu-Shen Liu, Hui Zhang, Yi Fang, Zhizhong Han: Retrospective Feature Pyramid Network for Point Cloud Semantic Segmentation. ICCV 2023
Mix3D	0.781 3	0.964 2	0.855 1	0.843 16	0.781 7	0.858 12	0.575 6	0.831 32	0.685 14	0.714 2	0.979 1	0.594 8	0.310 26	0.801 1	0.892 16	0.841 2	0.819 4	0.723 4	0.940 13	0.887 6	0.725 23
Alexey Nekrasov, Jonas Schult, Or Litany, Bastian Leibe, Francis Engelmann: Mix3D: Out-of-Context Data Augmentation for 3D Scenes. 3DV 2021 (Oral)
VMNet	0.746 22	0.870 18	0.838 2	0.858 4	0.729 30	0.850 20	0.501 36	0.874 15	0.587 53	0.658 18	0.956 25	0.564 22	0.299 30	0.765 15	0.900 12	0.716 36	0.812 11	0.631 38	0.939 14	0.858 27	0.709 30
Zeyu HU, Xuyang Bai, Jiaxiang Shang, Runze Zhang, Jiayu Dong, Xin Wang, Guangyuan Sun, Hongbo Fu, Chiew-Lan Tai: VMNet: Voxel-Mesh Network for Geodesic-Aware 3D Semantic Segmentation. ICCV 2021 (Oral)
ResLFE_HDS	0.772 6	0.939 4	0.824 6	0.854 6	0.771 10	0.840 30	0.564 10	0.900 8	0.686 13	0.677 11	0.961 16	0.537 30	0.348 11	0.769 11	0.903 10	0.785 10	0.815 6	0.676 21	0.939 14	0.880 11	0.772 8

MatchingNet	0.724 34	0.812 37	0.812 19	0.810 33	0.735 28	0.834 37	0.495 42	0.860 22	0.572 60	0.602 43	0.954 34	0.512 41	0.280 42	0.757 20	0.845 37	0.725 30	0.780 34	0.606 49	0.937 16	0.851 35	0.700 34

OA-CNN-L_ScanNet20	0.756 13	0.783 44	0.826 5	0.858 4	0.776 8	0.837 33	0.548 15	0.896 11	0.649 25	0.675 12	0.962 15	0.586 13	0.335 17	0.771 10	0.802 48	0.770 15	0.787 32	0.691 14	0.936 17	0.880 11	0.761 10

Virtual MVFusion	0.746 22	0.771 52	0.819 12	0.848 11	0.702 36	0.865 9	0.397 84	0.899 9	0.699 6	0.664 17	0.948 54	0.588 11	0.330 19	0.746 27	0.851 35	0.764 17	0.796 23	0.704 9	0.935 18	0.866 22	0.728 19
Abhijit Kundu, Xiaoqi Yin, Alireza Fathi, David Ross, Brian Brewington, Thomas Funkhouser, Caroline Pantofaru: Virtual Multi-view Fusion for 3D Semantic Segmentation. ECCV 2020
SparseConvNet	0.725 32	0.647 89	0.821 9	0.846 13	0.721 32	0.869 6	0.533 22	0.754 57	0.603 46	0.614 35	0.955 28	0.572 19	0.325 21	0.710 33	0.870 21	0.724 31	0.823 2	0.628 39	0.934 19	0.865 23	0.683 38

INS-Conv-semantic	0.717 35	0.751 61	0.759 51	0.812 31	0.704 35	0.868 7	0.537 21	0.842 28	0.609 42	0.608 39	0.953 38	0.534 33	0.293 33	0.616 52	0.864 24	0.719 35	0.793 27	0.640 34	0.933 20	0.845 40	0.663 44

LargeKernel3D	0.739 28	0.909 10	0.820 10	0.806 38	0.740 26	0.852 18	0.545 16	0.826 34	0.594 51	0.643 23	0.955 28	0.541 29	0.263 56	0.723 32	0.858 28	0.775 14	0.767 42	0.678 18	0.933 20	0.848 36	0.694 35
Yukang Chen, Jianhui Liu, Xiangyu Zhang, Xiaojuan Qi, Jiaya Jia: LargeKernel3D: Scaling up Kernels in 3D Sparse CNNs. CVPR 2023
PointTransformer++	0.725 32	0.727 74	0.811 20	0.819 27	0.765 13	0.841 29	0.502 35	0.814 41	0.621 36	0.623 33	0.955 28	0.556 25	0.284 39	0.620 51	0.866 23	0.781 11	0.757 53	0.648 30	0.932 22	0.862 24	0.709 30

O-CNN	0.762 11	0.924 8	0.823 7	0.844 15	0.770 11	0.852 18	0.577 4	0.847 27	0.711 2	0.640 27	0.958 20	0.592 9	0.217 72	0.762 16	0.888 17	0.758 19	0.813 10	0.726 2	0.932 22	0.868 20	0.744 13
Peng-Shuai Wang, Yang Liu, Yu-Xiao Guo, Chun-Yu Sun, Xin Tong: O-CNN: Octree-based Convolutional Neural Networks for 3D Shape Analysis. SIGGRAPH 2017
RPN	0.736 29	0.776 48	0.790 35	0.851 7	0.754 19	0.854 14	0.491 45	0.866 19	0.596 50	0.686 7	0.955 28	0.536 31	0.342 13	0.624 49	0.869 22	0.787 9	0.802 15	0.628 39	0.927 24	0.875 17	0.704 32

CU-Hybrid Net	0.764 9	0.924 8	0.819 12	0.840 17	0.757 17	0.853 16	0.580 3	0.848 25	0.709 3	0.643 23	0.958 20	0.587 12	0.295 32	0.753 23	0.884 20	0.758 19	0.815 6	0.725 3	0.927 24	0.867 21	0.743 14

dtc_net	0.625 80	0.703 80	0.751 57	0.794 44	0.535 84	0.848 23	0.480 47	0.676 79	0.528 73	0.469 82	0.944 68	0.454 63	0.004 113	0.464 90	0.636 75	0.704 42	0.758 51	0.548 79	0.924 26	0.787 76	0.492 96

OctFormer	0.766 7	0.925 7	0.808 22	0.849 9	0.786 5	0.846 26	0.566 9	0.876 14	0.690 10	0.674 13	0.960 17	0.576 17	0.226 66	0.753 23	0.904 9	0.777 12	0.815 6	0.722 5	0.923 27	0.877 13	0.776 7
Peng-Shuai Wang: OctFormer: Octree-based Transformers for 3D Point Clouds. SIGGRAPH 2023
SAT	0.742 26	0.860 21	0.765 48	0.819 27	0.769 12	0.848 23	0.533 22	0.829 33	0.663 20	0.631 30	0.955 28	0.586 13	0.274 46	0.753 23	0.896 14	0.729 27	0.760 49	0.666 27	0.921 28	0.855 31	0.733 17

MSP	0.748 20	0.623 93	0.804 24	0.859 3	0.745 25	0.824 47	0.501 36	0.912 4	0.690 10	0.685 8	0.956 25	0.567 20	0.320 23	0.768 13	0.918 5	0.720 33	0.802 15	0.676 21	0.921 28	0.881 10	0.779 6

DMF-Net	0.752 16	0.906 12	0.793 34	0.802 40	0.689 39	0.825 45	0.556 12	0.867 17	0.681 15	0.602 43	0.960 17	0.555 26	0.365 7	0.779 5	0.859 26	0.747 22	0.795 26	0.717 6	0.917 30	0.856 29	0.764 9
C.Yang, Y.Yan, W.Zhao, J.Ye, X.Yang, A.Hussain, B.Dong, K.Huang: Towards Deeper and Better Multi-view Feature Fusion for 3D Semantic Segmentation. ICONIP 2023
BPNet	0.749 18	0.909 10	0.818 14	0.811 32	0.752 20	0.839 32	0.485 46	0.842 28	0.673 17	0.644 22	0.957 24	0.528 36	0.305 28	0.773 9	0.859 26	0.788 8	0.818 5	0.693 13	0.916 31	0.856 29	0.723 24
Wenbo Hu, Hengshuang Zhao, Li Jiang, Jiaya Jia, Tien-Tsin Wong: Bidirectional Projection Network for Cross Dimension Scene Understanding. CVPR 2021 (Oral)
Swin3D	0.779 4	0.861 20	0.818 14	0.836 19	0.790 3	0.875 4	0.576 5	0.905 6	0.704 5	0.739 1	0.969 10	0.611 2	0.349 10	0.756 21	0.958 1	0.702 44	0.805 14	0.708 7	0.916 31	0.898 3	0.801 2

LRPNet	0.742 26	0.816 35	0.806 23	0.807 36	0.752 20	0.828 43	0.575 6	0.839 30	0.699 6	0.637 28	0.954 34	0.520 39	0.320 23	0.755 22	0.834 39	0.760 18	0.772 38	0.676 21	0.915 33	0.862 24	0.717 26

PointContrast_LA_SEM	0.683 50	0.757 59	0.784 38	0.786 46	0.639 56	0.824 47	0.408 79	0.775 50	0.604 45	0.541 59	0.934 86	0.532 34	0.269 52	0.552 70	0.777 50	0.645 70	0.793 27	0.640 34	0.913 34	0.824 47	0.671 41

Feature-Geometry Net	0.685 46	0.866 19	0.748 59	0.819 27	0.645 54	0.794 71	0.450 62	0.802 45	0.587 53	0.604 41	0.945 62	0.464 58	0.201 80	0.554 69	0.840 38	0.723 32	0.732 63	0.602 52	0.907 35	0.822 50	0.603 66

EQ-Net	0.743 25	0.620 94	0.799 29	0.849 9	0.730 29	0.822 49	0.493 43	0.897 10	0.664 19	0.681 9	0.955 28	0.562 23	0.378 3	0.760 17	0.903 10	0.738 24	0.801 19	0.673 25	0.907 35	0.877 13	0.745 12
Zetong Yang, Li Jiang, Yanan Sun, Bernt Schiele, Jiaya JIa: A Unified Query-based Paradigm for Point Cloud Understanding. CVPR 2022
Feature_GeometricNet	0.690 44	0.884 16	0.754 55	0.795 43	0.647 52	0.818 56	0.422 76	0.802 45	0.612 39	0.604 41	0.945 62	0.462 59	0.189 85	0.563 66	0.853 33	0.726 29	0.765 44	0.632 37	0.904 37	0.821 51	0.606 63
Kangcheng Liu, Ben M. Chen: https://arxiv.org/abs/2012.09439. arXiv Preprint
AttAN	0.609 84	0.760 57	0.667 91	0.649 87	0.521 87	0.793 72	0.457 57	0.648 83	0.528 73	0.434 92	0.947 56	0.401 86	0.153 98	0.454 91	0.721 61	0.648 66	0.717 67	0.536 85	0.904 37	0.765 85	0.485 98
Gege Zhang, Qinghua Ma, Licheng Jiao, Fang Liu and Qigong Sun: AttAN: Attention Adversarial Networks for 3D Point Cloud Semantic Segmentation. IJCAI2020
PicassoNet-II	0.692 43	0.732 70	0.772 43	0.786 46	0.677 43	0.866 8	0.517 29	0.848 25	0.509 79	0.626 31	0.952 42	0.536 31	0.225 68	0.545 73	0.704 64	0.689 51	0.810 12	0.564 69	0.903 39	0.854 33	0.729 18
Huan Lei, Naveed Akhtar, Mubarak Shah, and Ajmal Mian: Geometric feature learning for 3D meshes.
PointConv	0.666 56	0.781 45	0.759 51	0.699 69	0.644 55	0.822 49	0.475 48	0.779 49	0.564 65	0.504 75	0.953 38	0.428 76	0.203 79	0.586 58	0.754 53	0.661 60	0.753 54	0.588 60	0.902 40	0.813 56	0.642 51
Wenxuan Wu, Zhongang Qi, Li Fuxin: PointConv: Deep Convolutional Networks on 3D Point Clouds. CVPR 2019
One Thing One Click	0.701 40	0.825 32	0.796 30	0.723 61	0.716 33	0.832 39	0.433 74	0.816 38	0.634 31	0.609 38	0.969 10	0.418 82	0.344 12	0.559 67	0.833 40	0.715 37	0.808 13	0.560 70	0.902 40	0.847 37	0.680 39

PointNet2-SFPN	0.631 74	0.771 52	0.692 84	0.672 77	0.524 86	0.837 33	0.440 71	0.706 73	0.538 70	0.446 87	0.944 68	0.421 81	0.219 71	0.552 70	0.751 55	0.591 85	0.737 60	0.543 82	0.901 42	0.768 84	0.557 83

StratifiedFormer	0.747 21	0.901 13	0.803 25	0.845 14	0.757 17	0.846 26	0.512 31	0.825 35	0.696 8	0.645 21	0.956 25	0.576 17	0.262 57	0.744 28	0.861 25	0.742 23	0.770 41	0.705 8	0.899 43	0.860 26	0.734 16
Xin Lai, Jianhui Liu, Li Jiang, Liwei Wang, Hengshuang Zhao, Shu Liu, Xiaojuan Qi, Jiaya Jia: Stratified Transformer for 3D Point Cloud Segmentation. CVPR 2022
HPEIN	0.618 82	0.729 72	0.668 90	0.647 88	0.597 67	0.766 84	0.414 78	0.680 76	0.520 75	0.525 66	0.946 59	0.432 72	0.215 73	0.493 86	0.599 79	0.638 71	0.617 97	0.570 64	0.897 44	0.806 58	0.605 65
Li Jiang, Hengshuang Zhao, Shu Liu, Xiaoyong Shen, Chi-Wing Fu, Jiaya Jia: Hierarchical Point-Edge Interaction Network for Point Cloud Semantic Segmentation. ICCV 2019
FusionAwareConv	0.630 77	0.604 97	0.741 66	0.766 55	0.590 69	0.747 90	0.501 36	0.734 64	0.503 81	0.527 65	0.919 96	0.454 63	0.323 22	0.550 72	0.420 97	0.678 53	0.688 79	0.544 80	0.896 45	0.795 66	0.627 57
Jiazhao Zhang, Chenyang Zhu, Lintao Zheng, Kai Xu: Fusion-Aware Point Convolution for Online Semantic 3D Scene Segmentation. CVPR 2020
Superpoint Network	0.683 50	0.851 24	0.728 70	0.800 42	0.653 49	0.806 64	0.468 52	0.804 43	0.572 60	0.602 43	0.946 59	0.453 66	0.239 65	0.519 78	0.822 42	0.689 51	0.762 48	0.595 56	0.895 46	0.827 46	0.630 56

PointMRNet	0.640 66	0.717 77	0.701 77	0.692 72	0.576 75	0.801 67	0.467 54	0.716 68	0.563 66	0.459 85	0.953 38	0.429 75	0.169 92	0.581 59	0.854 32	0.605 79	0.710 68	0.550 77	0.894 47	0.793 69	0.575 74

SALANet	0.670 54	0.816 35	0.770 46	0.768 53	0.652 50	0.807 63	0.451 59	0.747 59	0.659 23	0.545 58	0.924 92	0.473 55	0.149 100	0.571 63	0.811 46	0.635 73	0.746 58	0.623 42	0.892 48	0.794 67	0.570 76

RFCR	0.702 39	0.889 15	0.745 62	0.813 30	0.672 44	0.818 56	0.493 43	0.815 40	0.623 34	0.610 37	0.947 56	0.470 56	0.249 61	0.594 55	0.848 36	0.705 41	0.779 35	0.646 31	0.892 48	0.823 48	0.611 59
Jingyu Gong, Jiachen Xu, Xin Tan, Haichuan Song, Yanyun Qu, Yuan Xie, Lizhuang Ma: Omni-Supervised Point Cloud Segmentation via Gradual Receptive Field Component Reasoning. CVPR2021
contrastBoundary	0.705 37	0.769 55	0.775 42	0.809 34	0.687 40	0.820 52	0.439 72	0.812 42	0.661 21	0.591 49	0.945 62	0.515 40	0.171 90	0.633 46	0.856 29	0.720 33	0.796 23	0.668 26	0.889 50	0.847 37	0.689 36
Liyao Tang, Yibing Zhan, Zhe Chen, Baosheng Yu, Dacheng Tao: Contrastive Boundary Learning for Point Cloud Segmentation. CVPR2022
VI-PointConv	0.676 52	0.770 54	0.754 55	0.783 49	0.621 60	0.814 59	0.552 13	0.758 55	0.571 62	0.557 55	0.954 34	0.529 35	0.268 54	0.530 76	0.682 68	0.675 54	0.719 66	0.603 51	0.888 51	0.833 42	0.665 43
Xingyi Li, Wenxuan Wu, Xiaoli Z. Fern, Li Fuxin: The Devils in the Point Clouds: Studying the Robustness of Point Cloud Convolutions.
SConv	0.636 70	0.830 30	0.697 80	0.752 59	0.572 77	0.780 79	0.445 66	0.716 68	0.529 72	0.530 64	0.951 44	0.446 70	0.170 91	0.507 83	0.666 72	0.636 72	0.682 81	0.541 83	0.886 52	0.799 62	0.594 70

DTC	0.757 12	0.843 26	0.820 10	0.847 12	0.791 2	0.862 10	0.511 32	0.870 16	0.707 4	0.652 19	0.954 34	0.604 6	0.279 43	0.760 17	0.942 2	0.734 26	0.766 43	0.701 10	0.884 53	0.874 18	0.736 15

ROSMRF3D	0.673 53	0.789 41	0.748 59	0.763 56	0.635 58	0.814 59	0.407 81	0.747 59	0.581 57	0.573 52	0.950 48	0.484 50	0.271 50	0.607 53	0.754 53	0.649 64	0.774 37	0.596 54	0.883 54	0.823 48	0.606 63

KP-FCNN	0.684 47	0.847 25	0.758 53	0.784 48	0.647 52	0.814 59	0.473 49	0.772 51	0.605 44	0.594 48	0.935 82	0.450 67	0.181 88	0.587 56	0.805 47	0.690 49	0.785 33	0.614 45	0.882 55	0.819 52	0.632 55
H. Thomas, C. Qi, J. Deschaud, B. Marcotegui, F. Goulette, L. Guibas.: KPConv: Flexible and Deformable Convolution for Point Clouds. ICCV 2019
VACNN++	0.684 47	0.728 73	0.757 54	0.776 51	0.690 37	0.804 66	0.464 55	0.816 38	0.577 59	0.587 50	0.945 62	0.508 43	0.276 45	0.671 36	0.710 62	0.663 59	0.750 57	0.589 59	0.881 56	0.832 44	0.653 47

DenSeR	0.628 78	0.800 38	0.625 100	0.719 64	0.545 83	0.806 64	0.445 66	0.597 89	0.448 96	0.519 70	0.938 78	0.481 51	0.328 20	0.489 87	0.499 91	0.657 62	0.759 50	0.592 57	0.881 56	0.797 65	0.634 54

ConDaFormer	0.755 14	0.927 6	0.822 8	0.836 19	0.801 1	0.849 21	0.516 30	0.864 21	0.651 24	0.680 10	0.958 20	0.584 15	0.282 40	0.759 19	0.855 31	0.728 28	0.802 15	0.678 18	0.880 58	0.873 19	0.756 11
Lunhao Duan, Shanshan Zhao, Nan Xue, Mingming Gong, Guisong Xia, Dacheng Tao: ConDaFormer : Disassembled Transformer with Local Structure Enhancement for 3D Point Cloud Understanding. Neurips, 2023
ClickSeg_Semantic	0.703 38	0.774 50	0.800 26	0.793 45	0.760 14	0.847 25	0.471 50	0.802 45	0.463 93	0.634 29	0.968 12	0.491 47	0.271 50	0.726 31	0.910 7	0.706 40	0.815 6	0.551 76	0.878 59	0.833 42	0.570 76

MVPNet	0.641 64	0.831 29	0.715 71	0.671 79	0.590 69	0.781 77	0.394 85	0.679 77	0.642 27	0.553 56	0.937 79	0.462 59	0.256 58	0.649 40	0.406 98	0.626 74	0.691 78	0.666 27	0.877 60	0.792 72	0.608 62
Maximilian Jaritz, Jiayuan Gu, Hao Su: Multi-view PointNet for 3D Scene Understanding. GMDL Workshop, ICCV 2019
SIConv	0.625 80	0.830 30	0.694 82	0.757 57	0.563 79	0.772 83	0.448 63	0.647 84	0.520 75	0.509 71	0.949 52	0.431 74	0.191 84	0.496 85	0.614 78	0.647 67	0.672 85	0.535 86	0.876 61	0.783 78	0.571 75

PointConv-SFPN	0.641 64	0.776 48	0.703 75	0.721 63	0.557 81	0.826 44	0.451 59	0.672 80	0.563 66	0.483 78	0.943 71	0.425 79	0.162 95	0.644 42	0.726 58	0.659 61	0.709 70	0.572 63	0.875 62	0.786 77	0.559 82

MVF-GNN	0.658 59	0.558 101	0.751 57	0.655 84	0.690 37	0.722 93	0.453 58	0.867 17	0.579 58	0.576 51	0.893 104	0.523 37	0.293 33	0.733 29	0.571 82	0.692 46	0.659 88	0.606 49	0.875 62	0.804 60	0.668 42

3DSM_DMMF	0.631 74	0.626 92	0.745 62	0.801 41	0.607 63	0.751 89	0.506 33	0.729 66	0.565 64	0.491 77	0.866 107	0.434 71	0.197 83	0.595 54	0.630 76	0.709 39	0.705 73	0.560 70	0.875 62	0.740 92	0.491 97

MinkowskiNet	0.736 29	0.859 22	0.818 14	0.832 23	0.709 34	0.840 30	0.521 28	0.853 23	0.660 22	0.643 23	0.951 44	0.544 28	0.286 38	0.731 30	0.893 15	0.675 54	0.772 38	0.683 17	0.874 65	0.852 34	0.727 21
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019
SegGroup_sem	0.627 79	0.818 34	0.747 61	0.701 68	0.602 65	0.764 85	0.385 90	0.629 86	0.490 84	0.508 72	0.931 89	0.409 84	0.201 80	0.564 65	0.725 59	0.618 76	0.692 77	0.539 84	0.873 66	0.794 67	0.548 86
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
APCF-Net	0.631 74	0.742 66	0.687 89	0.672 77	0.557 81	0.792 74	0.408 79	0.665 81	0.545 69	0.508 72	0.952 42	0.428 76	0.186 86	0.634 45	0.702 65	0.620 75	0.706 72	0.555 74	0.873 66	0.798 64	0.581 72
Haojia, Lin: Adaptive Pyramid Context Fusion for Point Cloud Perception. GRSL
FPConv	0.639 67	0.785 43	0.760 50	0.713 67	0.603 64	0.798 69	0.392 86	0.534 100	0.603 46	0.524 67	0.948 54	0.457 61	0.250 60	0.538 74	0.723 60	0.598 83	0.696 76	0.614 45	0.872 68	0.799 62	0.567 79
Yiqun Lin, Zizheng Yan, Haibin Huang, Dong Du, Ligang Liu, Shuguang Cui, Xiaoguang Han: FPConv: Learning Local Flattening for Point Convolution. CVPR 2020
JSENet	0.699 41	0.881 17	0.762 49	0.821 25	0.667 45	0.800 68	0.522 27	0.792 48	0.613 38	0.607 40	0.935 82	0.492 46	0.205 77	0.576 60	0.853 33	0.691 48	0.758 51	0.652 29	0.872 68	0.828 45	0.649 48
Zeyu HU, Mingmin Zhen, Xuyang BAI, Hongbo Fu, Chiew-lan Tai: JSENet: Joint Semantic Segmentation and Edge Detection Network for 3D Point Clouds. ECCV 2020
CCRFNet	0.589 88	0.766 56	0.659 95	0.683 74	0.470 96	0.740 92	0.387 89	0.620 88	0.490 84	0.476 80	0.922 94	0.355 95	0.245 63	0.511 79	0.511 89	0.571 90	0.643 93	0.493 94	0.872 68	0.762 86	0.600 67

DGNet	0.684 47	0.712 78	0.784 38	0.782 50	0.658 46	0.835 36	0.499 40	0.823 37	0.641 28	0.597 46	0.950 48	0.487 49	0.281 41	0.575 61	0.619 77	0.647 67	0.764 45	0.620 44	0.871 71	0.846 39	0.688 37

PPCNN++	0.663 58	0.746 63	0.708 73	0.722 62	0.638 57	0.820 52	0.451 59	0.566 95	0.599 48	0.541 59	0.950 48	0.510 42	0.313 25	0.648 41	0.819 44	0.616 78	0.682 81	0.590 58	0.869 72	0.810 57	0.656 46
Pyunghwan Ahn, Juyoung Yang, Eojindl Yi, Chanho Lee, Junmo Kim: Projection-based Point Convolution for Efficient Point Cloud Segmentation. IEEE Access
PointMTL	0.632 73	0.731 71	0.688 87	0.675 76	0.591 68	0.784 76	0.444 69	0.565 96	0.610 40	0.492 76	0.949 52	0.456 62	0.254 59	0.587 56	0.706 63	0.599 82	0.665 87	0.612 48	0.868 73	0.791 75	0.579 73

PD-Net	0.638 68	0.797 39	0.769 47	0.641 92	0.590 69	0.820 52	0.461 56	0.537 99	0.637 30	0.536 62	0.947 56	0.388 89	0.206 76	0.656 38	0.668 71	0.647 67	0.732 63	0.585 61	0.868 73	0.793 69	0.473 102

Supervoxel-CNN	0.635 71	0.656 87	0.711 72	0.719 64	0.613 62	0.757 88	0.444 69	0.765 53	0.534 71	0.566 53	0.928 90	0.478 53	0.272 48	0.636 43	0.531 86	0.664 58	0.645 92	0.508 90	0.864 75	0.792 72	0.611 59

FusionNet	0.688 45	0.704 79	0.741 66	0.754 58	0.656 47	0.829 41	0.501 36	0.741 62	0.609 42	0.548 57	0.950 48	0.522 38	0.371 4	0.633 46	0.756 52	0.715 37	0.771 40	0.623 42	0.861 76	0.814 54	0.658 45
Feihu Zhang, Jin Fang, Benjamin Wah, Philip Torr: Deep FusionNet for Point Cloud Semantic Segmentation. ECCV 2020
SPH3D-GCN	0.610 83	0.858 23	0.772 43	0.489 105	0.532 85	0.792 74	0.404 82	0.643 85	0.570 63	0.507 74	0.935 82	0.414 83	0.046 110	0.510 80	0.702 65	0.602 81	0.705 73	0.549 78	0.859 77	0.773 83	0.534 89
Huan Lei, Naveed Akhtar, and Ajmal Mian: Spherical Kernel for Efficient Graph Convolution on 3D Point Clouds. TPAMI 2020
SAFNet-seg	0.654 62	0.752 60	0.734 68	0.664 82	0.583 73	0.815 58	0.399 83	0.754 57	0.639 29	0.535 63	0.942 72	0.470 56	0.309 27	0.665 37	0.539 84	0.650 63	0.708 71	0.635 36	0.857 78	0.793 69	0.642 51
Linqing Zhao, Jiwen Lu, Jie Zhou: Similarity-Aware Fusion Network for 3D Semantic Segmentation. IROS 2021
HPGCNN	0.656 61	0.698 82	0.743 64	0.650 86	0.564 78	0.820 52	0.505 34	0.758 55	0.631 32	0.479 79	0.945 62	0.480 52	0.226 66	0.572 62	0.774 51	0.690 49	0.735 61	0.614 45	0.853 79	0.776 82	0.597 69
Jisheng Dang, Qingyong Hu, Yulan Guo, Jun Yang: HPGCNN.
PointMetaBase	0.714 36	0.835 28	0.785 37	0.821 25	0.684 41	0.846 26	0.531 24	0.865 20	0.614 37	0.596 47	0.953 38	0.500 44	0.246 62	0.674 35	0.888 17	0.692 46	0.764 45	0.624 41	0.849 80	0.844 41	0.675 40

PointSPNet	0.637 69	0.734 69	0.692 84	0.714 66	0.576 75	0.797 70	0.446 64	0.743 61	0.598 49	0.437 90	0.942 72	0.403 85	0.150 99	0.626 48	0.800 49	0.649 64	0.697 75	0.557 73	0.846 81	0.777 81	0.563 80

joint point-based	0.634 72	0.614 95	0.778 41	0.667 81	0.633 59	0.825 45	0.420 77	0.804 43	0.467 91	0.561 54	0.951 44	0.494 45	0.291 35	0.566 64	0.458 93	0.579 89	0.764 45	0.559 72	0.838 82	0.814 54	0.598 68
Hung-Yueh Chiang, Yen-Liang Lin, Yueh-Cheng Liu, Winston H. Hsu: A Unified Point-Based Framework for 3D Segmentation. 3DV 2019
wsss-transformer	0.600 85	0.634 91	0.743 64	0.697 71	0.601 66	0.781 77	0.437 73	0.585 92	0.493 83	0.446 87	0.933 87	0.394 87	0.011 112	0.654 39	0.661 74	0.603 80	0.733 62	0.526 87	0.832 83	0.761 87	0.480 99

RandLA-Net	0.645 63	0.778 46	0.731 69	0.699 69	0.577 74	0.829 41	0.446 64	0.736 63	0.477 88	0.523 69	0.945 62	0.454 63	0.269 52	0.484 88	0.749 56	0.618 76	0.738 59	0.599 53	0.827 84	0.792 72	0.621 58

DCM-Net	0.658 59	0.778 46	0.702 76	0.806 38	0.619 61	0.813 62	0.468 52	0.693 75	0.494 82	0.524 67	0.941 74	0.449 68	0.298 31	0.510 80	0.821 43	0.675 54	0.727 65	0.568 67	0.826 85	0.803 61	0.637 53
Jonas Schult, Francis Engelmann, Theodora Kontogianni, Bastian Leibe: DualConvMesh-Net: Joint Geodesic and Euclidean Convolutions on 3D Meshes. CVPR 2020 [Oral]
Pointnet++ & Feature	0.557 94	0.735 68	0.661 94	0.686 73	0.491 92	0.744 91	0.392 86	0.539 98	0.451 95	0.375 98	0.946 59	0.376 91	0.205 77	0.403 97	0.356 101	0.553 92	0.643 93	0.497 92	0.824 86	0.756 88	0.515 93

PointASNL	0.666 56	0.703 80	0.781 40	0.751 60	0.655 48	0.830 40	0.471 50	0.769 52	0.474 89	0.537 61	0.951 44	0.475 54	0.279 43	0.635 44	0.698 67	0.675 54	0.751 55	0.553 75	0.816 87	0.806 58	0.703 33
Xu Yan, Chaoda Zheng, Zhen Li, Sheng Wang, Shuguang Cui: PointASNL: Robust Point Clouds Processing using Nonlocal Neural Networks with Adaptive Sampling. CVPR 2020
PCNN	0.498 100	0.559 100	0.644 98	0.560 101	0.420 101	0.711 96	0.229 108	0.414 103	0.436 98	0.352 102	0.941 74	0.324 99	0.155 97	0.238 108	0.387 100	0.493 96	0.529 107	0.509 88	0.813 88	0.751 90	0.504 95

PanopticFusion-label	0.529 96	0.491 107	0.688 87	0.604 96	0.386 102	0.632 105	0.225 110	0.705 74	0.434 99	0.293 106	0.815 108	0.348 96	0.241 64	0.499 84	0.669 70	0.507 95	0.649 90	0.442 104	0.796 89	0.602 109	0.561 81
Gaku Narita, Takashi Seno, Tomoya Ishikawa, Yohsuke Kaji: PanopticFusion: Online Volumetric Semantic Mapping at the Level of Stuff and Things. IROS 2019 (to appear)
TextureNet	0.566 92	0.672 86	0.664 92	0.671 79	0.494 91	0.719 94	0.445 66	0.678 78	0.411 102	0.396 95	0.935 82	0.356 94	0.225 68	0.412 96	0.535 85	0.565 91	0.636 96	0.464 98	0.794 90	0.680 102	0.568 78
Jingwei Huang, Haotian Zhang, Li Yi, Thomas Funkerhouser, Matthias Niessner, Leonidas Guibas: TextureNet: Consistent Local Parametrizations for Learning from High-Resolution Signals on Meshes. CVPR
Weakly-Openseg v3	0.489 101	0.749 62	0.664 92	0.646 90	0.496 90	0.559 110	0.122 113	0.577 93	0.257 113	0.364 99	0.805 109	0.198 111	0.096 105	0.510 80	0.496 92	0.361 108	0.563 101	0.359 111	0.777 91	0.644 106	0.532 91

LAP-D	0.594 86	0.720 75	0.692 84	0.637 93	0.456 97	0.773 82	0.391 88	0.730 65	0.587 53	0.445 89	0.940 76	0.381 90	0.288 36	0.434 94	0.453 95	0.591 85	0.649 90	0.581 62	0.777 91	0.749 91	0.610 61

DPC	0.592 87	0.720 75	0.700 78	0.602 97	0.480 93	0.762 87	0.380 91	0.713 71	0.585 56	0.437 90	0.940 76	0.369 92	0.288 36	0.434 94	0.509 90	0.590 87	0.639 95	0.567 68	0.772 93	0.755 89	0.592 71
Francis Engelmann, Theodora Kontogianni, Bastian Leibe: Dilated Point Convolutions: On the Receptive Field Size of Point Convolutions on 3D Point Clouds. ICRA 2020
PointCNN with RGB	0.458 103	0.577 99	0.611 103	0.356 113	0.321 110	0.715 95	0.299 102	0.376 107	0.328 109	0.319 104	0.944 68	0.285 103	0.164 94	0.216 111	0.229 106	0.484 98	0.545 105	0.456 100	0.755 94	0.709 98	0.475 101
Yangyan Li, Rui Bu, Mingchao Sun, Baoquan Chen: PointCNN. NeurIPS 2018
GMLPs	0.538 95	0.495 106	0.693 83	0.647 88	0.471 95	0.793 72	0.300 100	0.477 102	0.505 80	0.358 100	0.903 102	0.327 98	0.081 107	0.472 89	0.529 87	0.448 102	0.710 68	0.509 88	0.746 95	0.737 93	0.554 85

3DMV, FTSDF	0.501 99	0.558 101	0.608 105	0.424 111	0.478 94	0.690 97	0.246 106	0.586 91	0.468 90	0.450 86	0.911 98	0.394 87	0.160 96	0.438 92	0.212 108	0.432 103	0.541 106	0.475 97	0.742 96	0.727 95	0.477 100

SD-DETR	0.576 90	0.746 63	0.609 104	0.445 109	0.517 88	0.643 104	0.366 92	0.714 70	0.456 94	0.468 83	0.870 106	0.432 72	0.264 55	0.558 68	0.674 69	0.586 88	0.688 79	0.482 96	0.739 97	0.733 94	0.537 88

FCPN	0.447 104	0.679 83	0.604 106	0.578 100	0.380 103	0.682 99	0.291 103	0.106 113	0.483 87	0.258 111	0.920 95	0.258 107	0.025 111	0.231 110	0.325 102	0.480 99	0.560 103	0.463 99	0.725 98	0.666 104	0.231 113
Dario Rethage, Johanna Wald, Jürgen Sturm, Nassir Navab, Federico Tombari: Fully-Convolutional Point Networks for Large-Scale Point Clouds. ECCV 2018
ROSMRF	0.580 89	0.772 51	0.707 74	0.681 75	0.563 79	0.764 85	0.362 93	0.515 101	0.465 92	0.465 84	0.936 81	0.427 78	0.207 75	0.438 92	0.577 81	0.536 93	0.675 84	0.486 95	0.723 99	0.779 79	0.524 92

DVVNet	0.562 93	0.648 88	0.700 78	0.770 52	0.586 72	0.687 98	0.333 97	0.650 82	0.514 78	0.475 81	0.906 100	0.359 93	0.223 70	0.340 101	0.442 96	0.422 104	0.668 86	0.501 91	0.708 100	0.779 79	0.534 89

3DMV	0.484 102	0.484 108	0.538 109	0.643 91	0.424 100	0.606 109	0.310 98	0.574 94	0.433 100	0.378 96	0.796 110	0.301 101	0.214 74	0.537 75	0.208 109	0.472 101	0.507 110	0.413 107	0.693 101	0.602 109	0.539 87
Angela Dai, Matthias Niessner: 3DMV: Joint 3D-Multi-View Prediction for 3D Semantic Scene Segmentation. ECCV'18
SQN_0.1%	0.569 91	0.676 84	0.696 81	0.657 83	0.497 89	0.779 80	0.424 75	0.548 97	0.515 77	0.376 97	0.902 103	0.422 80	0.357 8	0.379 99	0.456 94	0.596 84	0.659 88	0.544 80	0.685 102	0.665 105	0.556 84

Online SegFusion	0.515 98	0.607 96	0.644 98	0.579 99	0.434 99	0.630 106	0.353 94	0.628 87	0.440 97	0.410 93	0.762 112	0.307 100	0.167 93	0.520 77	0.403 99	0.516 94	0.565 100	0.447 102	0.678 103	0.701 99	0.514 94
Davide Menini, Suryansh Kumar, Martin R. Oswald, Erik Sandstroem, Cristian Sminchisescu, Luc van Gool: A Real-Time Learning Framework for Joint 3D Reconstruction and Semantic Segmentation. Robotics and Automation Letters Submission
PNET2	0.442 106	0.548 103	0.548 108	0.597 98	0.363 106	0.628 107	0.300 100	0.292 108	0.374 104	0.307 105	0.881 105	0.268 106	0.186 86	0.238 108	0.204 110	0.407 105	0.506 111	0.449 101	0.667 104	0.620 108	0.462 103

SurfaceConvPF	0.442 106	0.505 105	0.622 102	0.380 112	0.342 108	0.654 101	0.227 109	0.397 105	0.367 105	0.276 108	0.924 92	0.240 108	0.198 82	0.359 100	0.262 104	0.366 106	0.581 98	0.435 105	0.640 105	0.668 103	0.398 104
Hao Pan, Shilin Liu, Yang Liu, Xin Tong: Convolutional Neural Networks on 3D Surfaces Using Parallel Frames.
DGCNN_reproduce	0.446 105	0.474 109	0.623 101	0.463 107	0.366 105	0.651 102	0.310 98	0.389 106	0.349 107	0.330 103	0.937 79	0.271 105	0.126 102	0.285 104	0.224 107	0.350 110	0.577 99	0.445 103	0.625 106	0.723 96	0.394 105
Yue Wang, Yongbin Sun, Ziwei Liu, Sanjay E. Sarma, Michael M. Bronstein, Justin M. Solomon: Dynamic Graph CNN for Learning on Point Clouds. TOG 2019
Tangent Convolutions	0.438 108	0.437 111	0.646 97	0.474 106	0.369 104	0.645 103	0.353 94	0.258 110	0.282 111	0.279 107	0.918 97	0.298 102	0.147 101	0.283 105	0.294 103	0.487 97	0.562 102	0.427 106	0.619 107	0.633 107	0.352 108
Maxim Tatarchenko, Jaesik Park, Vladlen Koltun, Qian-Yi Zhou: Tangent convolutions for dense prediction in 3d. CVPR 2018
subcloud_weak	0.516 97	0.676 84	0.591 107	0.609 94	0.442 98	0.774 81	0.335 96	0.597 89	0.422 101	0.357 101	0.932 88	0.341 97	0.094 106	0.298 103	0.528 88	0.473 100	0.676 83	0.495 93	0.602 108	0.721 97	0.349 109

SPLAT Net	0.393 110	0.472 110	0.511 110	0.606 95	0.311 111	0.656 100	0.245 107	0.405 104	0.328 109	0.197 112	0.927 91	0.227 110	0.000 115	0.001 115	0.249 105	0.271 113	0.510 108	0.383 110	0.593 109	0.699 100	0.267 111
Hang Su, Varun Jampani, Deqing Sun, Subhransu Maji, Evangelos Kalogerakis, Ming-Hsuan Yang, Jan Kautz: SPLATNet: Sparse Lattice Networks for Point Cloud Processing. CVPR 2018
ScanNet+FTSDF	0.383 111	0.297 113	0.491 111	0.432 110	0.358 107	0.612 108	0.274 104	0.116 112	0.411 102	0.265 109	0.904 101	0.229 109	0.079 108	0.250 106	0.185 111	0.320 111	0.510 108	0.385 109	0.548 110	0.597 112	0.394 105

PointNet++	0.339 112	0.584 98	0.478 112	0.458 108	0.256 113	0.360 114	0.250 105	0.247 111	0.278 112	0.261 110	0.677 113	0.183 112	0.117 103	0.212 112	0.145 113	0.364 107	0.346 114	0.232 114	0.548 110	0.523 113	0.252 112
Charles R. Qi, Li Yi, Hao Su, Leonidas J. Guibas: pointnet++: deep hierarchical feature learning on point sets in a metric space.
SSC-UNet	0.308 113	0.353 112	0.290 114	0.278 114	0.166 114	0.553 111	0.169 112	0.286 109	0.147 114	0.148 114	0.908 99	0.182 113	0.064 109	0.023 114	0.018 115	0.354 109	0.363 112	0.345 112	0.546 112	0.685 101	0.278 110

ScanNet	0.306 114	0.203 114	0.366 113	0.501 103	0.311 111	0.524 112	0.211 111	0.002 115	0.342 108	0.189 113	0.786 111	0.145 114	0.102 104	0.245 107	0.152 112	0.318 112	0.348 113	0.300 113	0.460 113	0.437 114	0.182 114
Angela Dai, Angel X. Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, Matthias Nießner: ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes. CVPR'17
3DWSSS	0.425 109	0.525 104	0.647 96	0.522 102	0.324 109	0.488 113	0.077 114	0.712 72	0.353 106	0.401 94	0.636 114	0.281 104	0.176 89	0.340 101	0.565 83	0.175 114	0.551 104	0.398 108	0.370 114	0.602 109	0.361 107

ERROR	0.054 115	0.000 115	0.041 115	0.172 115	0.030 115	0.062 115	0.001 115	0.035 114	0.004 115	0.051 115	0.143 115	0.019 115	0.003 114	0.041 113	0.050 114	0.003 115	0.054 115	0.018 115	0.005 115	0.264 115	0.082 115

3D Semantic Label Benchmark