3D Semantic Label Benchmark
The 3D semantic labeling task involves predicting a semantic labeling of a 3D scan mesh.
Evaluation and metricsOur 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 | Info | 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 24 | 0.833 21 | 0.788 3 | 0.853 14 | 0.545 15 | 0.910 4 | 0.713 1 | 0.705 4 | 0.979 1 | 0.596 5 | 0.390 1 | 0.769 10 | 0.832 39 | 0.821 3 | 0.792 28 | 0.730 1 | 0.975 1 | 0.897 3 | 0.785 3 | |
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. | ||||||||||||||||||||||
Mix3D | 0.781 3 | 0.964 2 | 0.855 1 | 0.843 15 | 0.781 6 | 0.858 10 | 0.575 5 | 0.831 30 | 0.685 12 | 0.714 2 | 0.979 1 | 0.594 6 | 0.310 25 | 0.801 1 | 0.892 14 | 0.841 2 | 0.819 3 | 0.723 4 | 0.940 12 | 0.887 5 | 0.725 21 | |
Alexey Nekrasov, Jonas Schult, Or Litany, Bastian Leibe, Francis Engelmann: Mix3D: Out-of-Context Data Augmentation for 3D Scenes. 3DV 2021 (Oral) | ||||||||||||||||||||||
PTv3 ScanNet | 0.794 1 | 0.941 3 | 0.813 16 | 0.851 7 | 0.782 5 | 0.890 2 | 0.597 1 | 0.916 1 | 0.696 7 | 0.713 3 | 0.979 1 | 0.635 1 | 0.384 2 | 0.793 2 | 0.907 6 | 0.821 3 | 0.790 29 | 0.696 10 | 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 | ||||||||||||||||||||||
ResLFE_HDS | 0.772 5 | 0.939 4 | 0.824 6 | 0.854 6 | 0.771 8 | 0.840 28 | 0.564 9 | 0.900 7 | 0.686 11 | 0.677 10 | 0.961 15 | 0.537 28 | 0.348 11 | 0.769 10 | 0.903 8 | 0.785 9 | 0.815 5 | 0.676 19 | 0.939 13 | 0.880 10 | 0.772 7 | |
PPT-SpUNet-Joint | 0.766 6 | 0.932 5 | 0.794 30 | 0.829 23 | 0.751 20 | 0.854 12 | 0.540 19 | 0.903 6 | 0.630 31 | 0.672 13 | 0.963 13 | 0.565 19 | 0.357 8 | 0.788 3 | 0.900 10 | 0.737 24 | 0.802 14 | 0.685 14 | 0.950 6 | 0.887 5 | 0.780 4 | |
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 | ||||||||||||||||||||||
ConDaFormer | 0.755 12 | 0.927 6 | 0.822 8 | 0.836 18 | 0.801 1 | 0.849 19 | 0.516 29 | 0.864 19 | 0.651 22 | 0.680 9 | 0.958 19 | 0.584 13 | 0.282 39 | 0.759 17 | 0.855 29 | 0.728 26 | 0.802 14 | 0.678 16 | 0.880 56 | 0.873 17 | 0.756 10 | |
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 | ||||||||||||||||||||||
OctFormer | 0.766 6 | 0.925 7 | 0.808 20 | 0.849 9 | 0.786 4 | 0.846 24 | 0.566 8 | 0.876 13 | 0.690 9 | 0.674 12 | 0.960 16 | 0.576 15 | 0.226 64 | 0.753 21 | 0.904 7 | 0.777 11 | 0.815 5 | 0.722 5 | 0.923 26 | 0.877 12 | 0.776 6 | |
Peng-Shuai Wang: OctFormer: Octree-based Transformers for 3D Point Clouds. SIGGRAPH 2023 | ||||||||||||||||||||||
O-CNN | 0.762 10 | 0.924 8 | 0.823 7 | 0.844 14 | 0.770 9 | 0.852 16 | 0.577 3 | 0.847 25 | 0.711 2 | 0.640 25 | 0.958 19 | 0.592 7 | 0.217 70 | 0.762 15 | 0.888 15 | 0.758 18 | 0.813 9 | 0.726 2 | 0.932 21 | 0.868 18 | 0.744 12 | |
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 | ||||||||||||||||||||||
CU-Hybrid Net | 0.764 8 | 0.924 8 | 0.819 11 | 0.840 16 | 0.757 15 | 0.853 14 | 0.580 2 | 0.848 23 | 0.709 3 | 0.643 21 | 0.958 19 | 0.587 10 | 0.295 31 | 0.753 21 | 0.884 18 | 0.758 18 | 0.815 5 | 0.725 3 | 0.927 23 | 0.867 19 | 0.743 13 | |
LargeKernel3D | 0.739 26 | 0.909 10 | 0.820 10 | 0.806 36 | 0.740 24 | 0.852 16 | 0.545 15 | 0.826 32 | 0.594 49 | 0.643 21 | 0.955 27 | 0.541 27 | 0.263 54 | 0.723 30 | 0.858 26 | 0.775 13 | 0.767 41 | 0.678 16 | 0.933 19 | 0.848 34 | 0.694 33 | |
Yukang Chen*, Jianhui Liu*, Xiangyu Zhang, Xiaojuan Qi, Jiaya Jia: LargeKernel3D: Scaling up Kernels in 3D Sparse CNNs. CVPR 2023 | ||||||||||||||||||||||
BPNet | 0.749 16 | 0.909 10 | 0.818 13 | 0.811 31 | 0.752 18 | 0.839 30 | 0.485 44 | 0.842 26 | 0.673 15 | 0.644 20 | 0.957 23 | 0.528 34 | 0.305 27 | 0.773 8 | 0.859 24 | 0.788 7 | 0.818 4 | 0.693 11 | 0.916 30 | 0.856 27 | 0.723 22 | |
Wenbo Hu, Hengshuang Zhao, Li Jiang, Jiaya Jia, Tien-Tsin Wong: Bidirectional Projection Network for Cross Dimension Scene Understanding. CVPR 2021 (Oral) | ||||||||||||||||||||||
DMF-Net | 0.752 14 | 0.906 12 | 0.793 32 | 0.802 38 | 0.689 37 | 0.825 43 | 0.556 11 | 0.867 15 | 0.681 13 | 0.602 41 | 0.960 16 | 0.555 24 | 0.365 7 | 0.779 5 | 0.859 24 | 0.747 21 | 0.795 25 | 0.717 6 | 0.917 29 | 0.856 27 | 0.764 8 | |
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 | ||||||||||||||||||||||
StratifiedFormer | 0.747 19 | 0.901 13 | 0.803 23 | 0.845 13 | 0.757 15 | 0.846 24 | 0.512 30 | 0.825 33 | 0.696 7 | 0.645 19 | 0.956 24 | 0.576 15 | 0.262 55 | 0.744 26 | 0.861 23 | 0.742 22 | 0.770 40 | 0.705 8 | 0.899 42 | 0.860 24 | 0.734 14 | |
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 | ||||||||||||||||||||||
IPCA | 0.731 29 | 0.890 14 | 0.837 3 | 0.864 2 | 0.726 29 | 0.873 4 | 0.530 24 | 0.824 34 | 0.489 84 | 0.647 18 | 0.978 4 | 0.609 4 | 0.336 15 | 0.624 47 | 0.733 55 | 0.758 18 | 0.776 35 | 0.570 62 | 0.949 7 | 0.877 12 | 0.728 17 | |
RFCR | 0.702 37 | 0.889 15 | 0.745 60 | 0.813 29 | 0.672 42 | 0.818 54 | 0.493 41 | 0.815 38 | 0.623 32 | 0.610 35 | 0.947 54 | 0.470 54 | 0.249 59 | 0.594 53 | 0.848 34 | 0.705 39 | 0.779 34 | 0.646 29 | 0.892 47 | 0.823 46 | 0.611 57 | |
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 | ||||||||||||||||||||||
Feature_GeometricNet | 0.690 42 | 0.884 16 | 0.754 53 | 0.795 41 | 0.647 50 | 0.818 54 | 0.422 74 | 0.802 43 | 0.612 37 | 0.604 39 | 0.945 60 | 0.462 57 | 0.189 83 | 0.563 64 | 0.853 31 | 0.726 27 | 0.765 42 | 0.632 35 | 0.904 36 | 0.821 49 | 0.606 61 | |
Kangcheng Liu, Ben M. Chen: https://arxiv.org/abs/2012.09439. arXiv Preprint | ||||||||||||||||||||||
JSENet | 0.699 39 | 0.881 17 | 0.762 47 | 0.821 24 | 0.667 43 | 0.800 66 | 0.522 26 | 0.792 46 | 0.613 36 | 0.607 38 | 0.935 80 | 0.492 44 | 0.205 75 | 0.576 58 | 0.853 31 | 0.691 46 | 0.758 49 | 0.652 27 | 0.872 66 | 0.828 43 | 0.649 46 | |
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 | ||||||||||||||||||||||
VMNet | 0.746 20 | 0.870 18 | 0.838 2 | 0.858 4 | 0.729 28 | 0.850 18 | 0.501 34 | 0.874 14 | 0.587 51 | 0.658 17 | 0.956 24 | 0.564 20 | 0.299 29 | 0.765 14 | 0.900 10 | 0.716 34 | 0.812 10 | 0.631 36 | 0.939 13 | 0.858 25 | 0.709 28 | |
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) | ||||||||||||||||||||||
Feature-Geometry Net | 0.685 44 | 0.866 19 | 0.748 57 | 0.819 26 | 0.645 52 | 0.794 69 | 0.450 60 | 0.802 43 | 0.587 51 | 0.604 39 | 0.945 60 | 0.464 56 | 0.201 78 | 0.554 67 | 0.840 36 | 0.723 30 | 0.732 61 | 0.602 50 | 0.907 34 | 0.822 48 | 0.603 64 | |
Swin3D | 0.779 4 | 0.861 20 | 0.818 13 | 0.836 18 | 0.790 2 | 0.875 3 | 0.576 4 | 0.905 5 | 0.704 4 | 0.739 1 | 0.969 9 | 0.611 2 | 0.349 10 | 0.756 19 | 0.958 1 | 0.702 42 | 0.805 13 | 0.708 7 | 0.916 30 | 0.898 2 | 0.801 2 | |
SAT | 0.742 24 | 0.860 21 | 0.765 46 | 0.819 26 | 0.769 10 | 0.848 21 | 0.533 21 | 0.829 31 | 0.663 18 | 0.631 28 | 0.955 27 | 0.586 11 | 0.274 44 | 0.753 21 | 0.896 12 | 0.729 25 | 0.760 47 | 0.666 25 | 0.921 27 | 0.855 29 | 0.733 15 | |
MinkowskiNet | 0.736 27 | 0.859 22 | 0.818 13 | 0.832 22 | 0.709 32 | 0.840 28 | 0.521 27 | 0.853 21 | 0.660 20 | 0.643 21 | 0.951 42 | 0.544 26 | 0.286 37 | 0.731 28 | 0.893 13 | 0.675 52 | 0.772 37 | 0.683 15 | 0.874 63 | 0.852 32 | 0.727 19 | |
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019 | ||||||||||||||||||||||
SPH3D-GCN | 0.610 81 | 0.858 23 | 0.772 41 | 0.489 103 | 0.532 83 | 0.792 72 | 0.404 80 | 0.643 83 | 0.570 61 | 0.507 72 | 0.935 80 | 0.414 81 | 0.046 108 | 0.510 78 | 0.702 63 | 0.602 79 | 0.705 71 | 0.549 76 | 0.859 75 | 0.773 81 | 0.534 87 | |
Huan Lei, Naveed Akhtar, and Ajmal Mian: Spherical Kernel for Efficient Graph Convolution on 3D Point Clouds. TPAMI 2020 | ||||||||||||||||||||||
Superpoint Network | 0.683 48 | 0.851 24 | 0.728 68 | 0.800 40 | 0.653 47 | 0.806 62 | 0.468 50 | 0.804 41 | 0.572 58 | 0.602 41 | 0.946 57 | 0.453 64 | 0.239 63 | 0.519 76 | 0.822 40 | 0.689 49 | 0.762 46 | 0.595 54 | 0.895 45 | 0.827 44 | 0.630 54 | |
KP-FCNN | 0.684 45 | 0.847 25 | 0.758 51 | 0.784 46 | 0.647 50 | 0.814 57 | 0.473 47 | 0.772 49 | 0.605 42 | 0.594 46 | 0.935 80 | 0.450 65 | 0.181 86 | 0.587 54 | 0.805 45 | 0.690 47 | 0.785 32 | 0.614 43 | 0.882 53 | 0.819 50 | 0.632 53 | |
H. Thomas, C. Qi, J. Deschaud, B. Marcotegui, F. Goulette, L. Guibas.: KPConv: Flexible and Deformable Convolution for Point Clouds. ICCV 2019 | ||||||||||||||||||||||
Retro-FPN | 0.744 22 | 0.842 26 | 0.800 24 | 0.767 52 | 0.740 24 | 0.836 33 | 0.541 17 | 0.914 2 | 0.672 16 | 0.626 29 | 0.958 19 | 0.552 25 | 0.272 46 | 0.777 6 | 0.886 17 | 0.696 43 | 0.801 18 | 0.674 22 | 0.941 11 | 0.858 25 | 0.717 24 | |
Peng Xiang*, Xin Wen*, Yu-Shen Liu, Hui Zhang, Yi Fang, Zhizhong Han: Retrospective Feature Pyramid Network for Point Cloud Semantic Segmentation. ICCV 2023 | ||||||||||||||||||||||
PointMetaBase | 0.714 34 | 0.835 27 | 0.785 35 | 0.821 24 | 0.684 39 | 0.846 24 | 0.531 23 | 0.865 18 | 0.614 35 | 0.596 45 | 0.953 36 | 0.500 42 | 0.246 60 | 0.674 33 | 0.888 15 | 0.692 44 | 0.764 43 | 0.624 39 | 0.849 78 | 0.844 39 | 0.675 38 | |
MVPNet | 0.641 62 | 0.831 28 | 0.715 69 | 0.671 77 | 0.590 67 | 0.781 75 | 0.394 83 | 0.679 75 | 0.642 25 | 0.553 54 | 0.937 77 | 0.462 57 | 0.256 56 | 0.649 38 | 0.406 96 | 0.626 72 | 0.691 76 | 0.666 25 | 0.877 58 | 0.792 70 | 0.608 60 | |
Maximilian Jaritz, Jiayuan Gu, Hao Su: Multi-view PointNet for 3D Scene Understanding. GMDL Workshop, ICCV 2019 | ||||||||||||||||||||||
SConv | 0.636 68 | 0.830 29 | 0.697 78 | 0.752 57 | 0.572 75 | 0.780 77 | 0.445 64 | 0.716 66 | 0.529 70 | 0.530 62 | 0.951 42 | 0.446 68 | 0.170 89 | 0.507 81 | 0.666 70 | 0.636 70 | 0.682 79 | 0.541 81 | 0.886 51 | 0.799 60 | 0.594 68 | |
SIConv | 0.625 78 | 0.830 29 | 0.694 80 | 0.757 55 | 0.563 77 | 0.772 81 | 0.448 61 | 0.647 82 | 0.520 73 | 0.509 69 | 0.949 50 | 0.431 72 | 0.191 82 | 0.496 83 | 0.614 76 | 0.647 65 | 0.672 83 | 0.535 84 | 0.876 59 | 0.783 76 | 0.571 73 | |
One Thing One Click | 0.701 38 | 0.825 31 | 0.796 28 | 0.723 59 | 0.716 31 | 0.832 37 | 0.433 72 | 0.816 36 | 0.634 29 | 0.609 36 | 0.969 9 | 0.418 80 | 0.344 12 | 0.559 65 | 0.833 38 | 0.715 35 | 0.808 12 | 0.560 68 | 0.902 39 | 0.847 35 | 0.680 37 | |
O3DSeg | 0.668 53 | 0.822 32 | 0.771 43 | 0.496 102 | 0.651 49 | 0.833 36 | 0.541 17 | 0.761 52 | 0.555 66 | 0.611 34 | 0.966 12 | 0.489 46 | 0.370 5 | 0.388 96 | 0.580 78 | 0.776 12 | 0.751 53 | 0.570 62 | 0.956 5 | 0.817 51 | 0.646 48 | |
SegGroup_sem | 0.627 77 | 0.818 33 | 0.747 59 | 0.701 66 | 0.602 63 | 0.764 83 | 0.385 88 | 0.629 84 | 0.490 82 | 0.508 70 | 0.931 87 | 0.409 82 | 0.201 78 | 0.564 63 | 0.725 57 | 0.618 74 | 0.692 75 | 0.539 82 | 0.873 64 | 0.794 65 | 0.548 84 | |
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022 | ||||||||||||||||||||||
SALANet | 0.670 52 | 0.816 34 | 0.770 44 | 0.768 51 | 0.652 48 | 0.807 61 | 0.451 57 | 0.747 57 | 0.659 21 | 0.545 56 | 0.924 90 | 0.473 53 | 0.149 98 | 0.571 61 | 0.811 44 | 0.635 71 | 0.746 56 | 0.623 40 | 0.892 47 | 0.794 65 | 0.570 74 | |
LRPNet | 0.742 24 | 0.816 34 | 0.806 21 | 0.807 34 | 0.752 18 | 0.828 41 | 0.575 5 | 0.839 28 | 0.699 5 | 0.637 26 | 0.954 33 | 0.520 37 | 0.320 22 | 0.755 20 | 0.834 37 | 0.760 17 | 0.772 37 | 0.676 19 | 0.915 32 | 0.862 22 | 0.717 24 | |
MatchingNet | 0.724 32 | 0.812 36 | 0.812 17 | 0.810 32 | 0.735 26 | 0.834 35 | 0.495 40 | 0.860 20 | 0.572 58 | 0.602 41 | 0.954 33 | 0.512 39 | 0.280 41 | 0.757 18 | 0.845 35 | 0.725 28 | 0.780 33 | 0.606 47 | 0.937 15 | 0.851 33 | 0.700 32 | |
DenSeR | 0.628 76 | 0.800 37 | 0.625 98 | 0.719 62 | 0.545 81 | 0.806 62 | 0.445 64 | 0.597 87 | 0.448 94 | 0.519 68 | 0.938 76 | 0.481 49 | 0.328 19 | 0.489 85 | 0.499 89 | 0.657 60 | 0.759 48 | 0.592 55 | 0.881 54 | 0.797 63 | 0.634 52 | |
PD-Net | 0.638 66 | 0.797 38 | 0.769 45 | 0.641 90 | 0.590 67 | 0.820 50 | 0.461 54 | 0.537 97 | 0.637 28 | 0.536 60 | 0.947 54 | 0.388 87 | 0.206 74 | 0.656 36 | 0.668 69 | 0.647 65 | 0.732 61 | 0.585 59 | 0.868 71 | 0.793 67 | 0.473 100 | |
PointConvFormer | 0.749 16 | 0.793 39 | 0.790 33 | 0.807 34 | 0.750 21 | 0.856 11 | 0.524 25 | 0.881 12 | 0.588 50 | 0.642 24 | 0.977 7 | 0.591 8 | 0.274 44 | 0.781 4 | 0.929 2 | 0.804 5 | 0.796 22 | 0.642 31 | 0.947 8 | 0.885 7 | 0.715 27 | |
Wenxuan Wu, Qi Shan, Li Fuxin: PointConvFormer: Revenge of the Point-based Convolution. | ||||||||||||||||||||||
ROSMRF3D | 0.673 51 | 0.789 40 | 0.748 57 | 0.763 54 | 0.635 56 | 0.814 57 | 0.407 79 | 0.747 57 | 0.581 55 | 0.573 50 | 0.950 46 | 0.484 48 | 0.271 48 | 0.607 51 | 0.754 51 | 0.649 62 | 0.774 36 | 0.596 52 | 0.883 52 | 0.823 46 | 0.606 61 | |
PNE | 0.755 12 | 0.786 41 | 0.835 4 | 0.834 20 | 0.758 13 | 0.849 19 | 0.570 7 | 0.836 29 | 0.648 24 | 0.668 15 | 0.978 4 | 0.581 14 | 0.367 6 | 0.683 32 | 0.856 27 | 0.804 5 | 0.801 18 | 0.678 16 | 0.961 4 | 0.889 4 | 0.716 26 | |
P. Hermosilla: Point Neighborhood Embeddings. | ||||||||||||||||||||||
FPConv | 0.639 65 | 0.785 42 | 0.760 48 | 0.713 65 | 0.603 62 | 0.798 67 | 0.392 84 | 0.534 98 | 0.603 44 | 0.524 65 | 0.948 52 | 0.457 59 | 0.250 58 | 0.538 72 | 0.723 58 | 0.598 81 | 0.696 74 | 0.614 43 | 0.872 66 | 0.799 60 | 0.567 77 | |
Yiqun Lin, Zizheng Yan, Haibin Huang, Dong Du, Ligang Liu, Shuguang Cui, Xiaoguang Han: FPConv: Learning Local Flattening for Point Convolution. CVPR 2020 | ||||||||||||||||||||||
OA-CNN-L_ScanNet20 | 0.756 11 | 0.783 43 | 0.826 5 | 0.858 4 | 0.776 7 | 0.837 31 | 0.548 14 | 0.896 10 | 0.649 23 | 0.675 11 | 0.962 14 | 0.586 11 | 0.335 16 | 0.771 9 | 0.802 46 | 0.770 14 | 0.787 31 | 0.691 12 | 0.936 16 | 0.880 10 | 0.761 9 | |
PointConv | 0.666 54 | 0.781 44 | 0.759 49 | 0.699 67 | 0.644 53 | 0.822 47 | 0.475 46 | 0.779 47 | 0.564 63 | 0.504 73 | 0.953 36 | 0.428 74 | 0.203 77 | 0.586 56 | 0.754 51 | 0.661 58 | 0.753 52 | 0.588 58 | 0.902 39 | 0.813 54 | 0.642 49 | |
Wenxuan Wu, Zhongang Qi, Li Fuxin: PointConv: Deep Convolutional Networks on 3D Point Clouds. CVPR 2019 | ||||||||||||||||||||||
DCM-Net | 0.658 57 | 0.778 45 | 0.702 74 | 0.806 36 | 0.619 59 | 0.813 60 | 0.468 50 | 0.693 73 | 0.494 80 | 0.524 65 | 0.941 72 | 0.449 66 | 0.298 30 | 0.510 78 | 0.821 41 | 0.675 52 | 0.727 63 | 0.568 65 | 0.826 83 | 0.803 59 | 0.637 51 | |
Jonas Schult*, Francis Engelmann*, Theodora Kontogianni, Bastian Leibe: DualConvMesh-Net: Joint Geodesic and Euclidean Convolutions on 3D Meshes. CVPR 2020 [Oral] | ||||||||||||||||||||||
RandLA-Net | 0.645 61 | 0.778 45 | 0.731 67 | 0.699 67 | 0.577 72 | 0.829 39 | 0.446 62 | 0.736 61 | 0.477 86 | 0.523 67 | 0.945 60 | 0.454 61 | 0.269 50 | 0.484 86 | 0.749 54 | 0.618 74 | 0.738 57 | 0.599 51 | 0.827 82 | 0.792 70 | 0.621 56 | |
PointConv-SFPN | 0.641 62 | 0.776 47 | 0.703 73 | 0.721 61 | 0.557 79 | 0.826 42 | 0.451 57 | 0.672 78 | 0.563 64 | 0.483 76 | 0.943 69 | 0.425 77 | 0.162 93 | 0.644 40 | 0.726 56 | 0.659 59 | 0.709 68 | 0.572 61 | 0.875 60 | 0.786 75 | 0.559 80 | |
RPN | 0.736 27 | 0.776 47 | 0.790 33 | 0.851 7 | 0.754 17 | 0.854 12 | 0.491 43 | 0.866 17 | 0.596 48 | 0.686 6 | 0.955 27 | 0.536 29 | 0.342 13 | 0.624 47 | 0.869 20 | 0.787 8 | 0.802 14 | 0.628 37 | 0.927 23 | 0.875 16 | 0.704 30 | |
ClickSeg_Semantic | 0.703 36 | 0.774 49 | 0.800 24 | 0.793 43 | 0.760 12 | 0.847 23 | 0.471 48 | 0.802 43 | 0.463 91 | 0.634 27 | 0.968 11 | 0.491 45 | 0.271 48 | 0.726 29 | 0.910 5 | 0.706 38 | 0.815 5 | 0.551 74 | 0.878 57 | 0.833 40 | 0.570 74 | |
ROSMRF | 0.580 87 | 0.772 50 | 0.707 72 | 0.681 73 | 0.563 77 | 0.764 83 | 0.362 91 | 0.515 99 | 0.465 90 | 0.465 82 | 0.936 79 | 0.427 76 | 0.207 73 | 0.438 90 | 0.577 79 | 0.536 91 | 0.675 82 | 0.486 93 | 0.723 97 | 0.779 77 | 0.524 90 | |
Virtual MVFusion | 0.746 20 | 0.771 51 | 0.819 11 | 0.848 11 | 0.702 34 | 0.865 8 | 0.397 82 | 0.899 8 | 0.699 5 | 0.664 16 | 0.948 52 | 0.588 9 | 0.330 18 | 0.746 25 | 0.851 33 | 0.764 16 | 0.796 22 | 0.704 9 | 0.935 17 | 0.866 20 | 0.728 17 | |
Abhijit Kundu, Xiaoqi Yin, Alireza Fathi, David Ross, Brian Brewington, Thomas Funkhouser, Caroline Pantofaru: Virtual Multi-view Fusion for 3D Semantic Segmentation. ECCV 2020 | ||||||||||||||||||||||
PointNet2-SFPN | 0.631 72 | 0.771 51 | 0.692 82 | 0.672 75 | 0.524 84 | 0.837 31 | 0.440 69 | 0.706 71 | 0.538 68 | 0.446 85 | 0.944 66 | 0.421 79 | 0.219 69 | 0.552 68 | 0.751 53 | 0.591 83 | 0.737 58 | 0.543 80 | 0.901 41 | 0.768 82 | 0.557 81 | |
VI-PointConv | 0.676 50 | 0.770 53 | 0.754 53 | 0.783 47 | 0.621 58 | 0.814 57 | 0.552 12 | 0.758 53 | 0.571 60 | 0.557 53 | 0.954 33 | 0.529 33 | 0.268 52 | 0.530 74 | 0.682 66 | 0.675 52 | 0.719 64 | 0.603 49 | 0.888 50 | 0.833 40 | 0.665 41 | |
Xingyi Li, Wenxuan Wu, Xiaoli Z. Fern, Li Fuxin: The Devils in the Point Clouds: Studying the Robustness of Point Cloud Convolutions. | ||||||||||||||||||||||
contrastBoundary | 0.705 35 | 0.769 54 | 0.775 40 | 0.809 33 | 0.687 38 | 0.820 50 | 0.439 70 | 0.812 40 | 0.661 19 | 0.591 47 | 0.945 60 | 0.515 38 | 0.171 88 | 0.633 44 | 0.856 27 | 0.720 31 | 0.796 22 | 0.668 24 | 0.889 49 | 0.847 35 | 0.689 34 | |
Liyao Tang, Yibing Zhan, Zhe Chen, Baosheng Yu, Dacheng Tao: Contrastive Boundary Learning for Point Cloud Segmentation. CVPR2022 | ||||||||||||||||||||||
CCRFNet | 0.589 86 | 0.766 55 | 0.659 93 | 0.683 72 | 0.470 94 | 0.740 90 | 0.387 87 | 0.620 86 | 0.490 82 | 0.476 78 | 0.922 92 | 0.355 93 | 0.245 61 | 0.511 77 | 0.511 87 | 0.571 88 | 0.643 91 | 0.493 92 | 0.872 66 | 0.762 84 | 0.600 65 | |
AttAN | 0.609 82 | 0.760 56 | 0.667 89 | 0.649 85 | 0.521 85 | 0.793 70 | 0.457 55 | 0.648 81 | 0.528 71 | 0.434 90 | 0.947 54 | 0.401 84 | 0.153 96 | 0.454 89 | 0.721 59 | 0.648 64 | 0.717 65 | 0.536 83 | 0.904 36 | 0.765 83 | 0.485 96 | |
Gege Zhang, Qinghua Ma, Licheng Jiao, Fang Liu and Qigong Sun: AttAN: Attention Adversarial Networks for 3D Point Cloud Semantic Segmentation. IJCAI2020 | ||||||||||||||||||||||
OccuSeg+Semantic | 0.764 8 | 0.758 57 | 0.796 28 | 0.839 17 | 0.746 22 | 0.907 1 | 0.562 10 | 0.850 22 | 0.680 14 | 0.672 13 | 0.978 4 | 0.610 3 | 0.335 16 | 0.777 6 | 0.819 42 | 0.847 1 | 0.830 1 | 0.691 12 | 0.972 2 | 0.885 7 | 0.727 19 | |
PointContrast_LA_SEM | 0.683 48 | 0.757 58 | 0.784 36 | 0.786 44 | 0.639 54 | 0.824 45 | 0.408 77 | 0.775 48 | 0.604 43 | 0.541 57 | 0.934 84 | 0.532 32 | 0.269 50 | 0.552 68 | 0.777 48 | 0.645 68 | 0.793 26 | 0.640 32 | 0.913 33 | 0.824 45 | 0.671 39 | |
SAFNet-seg | 0.654 60 | 0.752 59 | 0.734 66 | 0.664 80 | 0.583 71 | 0.815 56 | 0.399 81 | 0.754 55 | 0.639 27 | 0.535 61 | 0.942 70 | 0.470 54 | 0.309 26 | 0.665 35 | 0.539 82 | 0.650 61 | 0.708 69 | 0.635 34 | 0.857 76 | 0.793 67 | 0.642 49 | |
Linqing Zhao, Jiwen Lu, Jie Zhou: Similarity-Aware Fusion Network for 3D Semantic Segmentation. IROS 2021 | ||||||||||||||||||||||
INS-Conv-semantic | 0.717 33 | 0.751 60 | 0.759 49 | 0.812 30 | 0.704 33 | 0.868 6 | 0.537 20 | 0.842 26 | 0.609 40 | 0.608 37 | 0.953 36 | 0.534 31 | 0.293 32 | 0.616 50 | 0.864 22 | 0.719 33 | 0.793 26 | 0.640 32 | 0.933 19 | 0.845 38 | 0.663 42 | |
Weakly-Openseg v3 | 0.489 99 | 0.749 61 | 0.664 90 | 0.646 88 | 0.496 88 | 0.559 108 | 0.122 111 | 0.577 91 | 0.257 111 | 0.364 97 | 0.805 107 | 0.198 109 | 0.096 103 | 0.510 78 | 0.496 90 | 0.361 106 | 0.563 99 | 0.359 109 | 0.777 89 | 0.644 104 | 0.532 89 | |
PPCNN++ | 0.663 56 | 0.746 62 | 0.708 71 | 0.722 60 | 0.638 55 | 0.820 50 | 0.451 57 | 0.566 93 | 0.599 46 | 0.541 57 | 0.950 46 | 0.510 40 | 0.313 24 | 0.648 39 | 0.819 42 | 0.616 76 | 0.682 79 | 0.590 56 | 0.869 70 | 0.810 55 | 0.656 44 | |
Pyunghwan Ahn, Juyoung Yang, Eojindl Yi, Chanho Lee, Junmo Kim: Projection-based Point Convolution for Efficient Point Cloud Segmentation. IEEE Access | ||||||||||||||||||||||
SD-DETR | 0.576 88 | 0.746 62 | 0.609 102 | 0.445 107 | 0.517 86 | 0.643 102 | 0.366 90 | 0.714 68 | 0.456 92 | 0.468 81 | 0.870 104 | 0.432 70 | 0.264 53 | 0.558 66 | 0.674 67 | 0.586 86 | 0.688 77 | 0.482 94 | 0.739 95 | 0.733 92 | 0.537 86 | |
One-Thing-One-Click | 0.693 40 | 0.743 64 | 0.794 30 | 0.655 82 | 0.684 39 | 0.822 47 | 0.497 39 | 0.719 65 | 0.622 33 | 0.617 32 | 0.977 7 | 0.447 67 | 0.339 14 | 0.750 24 | 0.664 71 | 0.703 41 | 0.790 29 | 0.596 52 | 0.946 9 | 0.855 29 | 0.647 47 | |
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 14 | 0.742 65 | 0.809 19 | 0.872 1 | 0.758 13 | 0.860 9 | 0.552 12 | 0.891 11 | 0.610 38 | 0.687 5 | 0.960 16 | 0.559 22 | 0.304 28 | 0.766 13 | 0.926 3 | 0.767 15 | 0.797 21 | 0.644 30 | 0.942 10 | 0.876 15 | 0.722 23 | |
Xiaoyang Wu, Yixing Lao, Li Jiang, Xihui Liu, Hengshuang Zhao: Point Transformer V2: Grouped Vector Attention and Partition-based Pooling. NeurIPS 2022 | ||||||||||||||||||||||
APCF-Net | 0.631 72 | 0.742 65 | 0.687 87 | 0.672 75 | 0.557 79 | 0.792 72 | 0.408 77 | 0.665 79 | 0.545 67 | 0.508 70 | 0.952 40 | 0.428 74 | 0.186 84 | 0.634 43 | 0.702 63 | 0.620 73 | 0.706 70 | 0.555 72 | 0.873 64 | 0.798 62 | 0.581 70 | |
Haojia, Lin: Adaptive Pyramid Context Fusion for Point Cloud Perception. GRSL | ||||||||||||||||||||||
Pointnet++ & Feature | 0.557 92 | 0.735 67 | 0.661 92 | 0.686 71 | 0.491 90 | 0.744 89 | 0.392 84 | 0.539 96 | 0.451 93 | 0.375 96 | 0.946 57 | 0.376 89 | 0.205 75 | 0.403 95 | 0.356 99 | 0.553 90 | 0.643 91 | 0.497 90 | 0.824 84 | 0.756 86 | 0.515 91 | |
PointSPNet | 0.637 67 | 0.734 68 | 0.692 82 | 0.714 64 | 0.576 73 | 0.797 68 | 0.446 62 | 0.743 59 | 0.598 47 | 0.437 88 | 0.942 70 | 0.403 83 | 0.150 97 | 0.626 46 | 0.800 47 | 0.649 62 | 0.697 73 | 0.557 71 | 0.846 79 | 0.777 79 | 0.563 78 | |
PicassoNet-II | 0.692 41 | 0.732 69 | 0.772 41 | 0.786 44 | 0.677 41 | 0.866 7 | 0.517 28 | 0.848 23 | 0.509 77 | 0.626 29 | 0.952 40 | 0.536 29 | 0.225 66 | 0.545 71 | 0.704 62 | 0.689 49 | 0.810 11 | 0.564 67 | 0.903 38 | 0.854 31 | 0.729 16 | |
Huan Lei, Naveed Akhtar, Mubarak Shah, and Ajmal Mian: Geometric feature learning for 3D meshes. | ||||||||||||||||||||||
PointMTL | 0.632 71 | 0.731 70 | 0.688 85 | 0.675 74 | 0.591 66 | 0.784 74 | 0.444 67 | 0.565 94 | 0.610 38 | 0.492 74 | 0.949 50 | 0.456 60 | 0.254 57 | 0.587 54 | 0.706 61 | 0.599 80 | 0.665 85 | 0.612 46 | 0.868 71 | 0.791 73 | 0.579 71 | |
HPEIN | 0.618 80 | 0.729 71 | 0.668 88 | 0.647 86 | 0.597 65 | 0.766 82 | 0.414 76 | 0.680 74 | 0.520 73 | 0.525 64 | 0.946 57 | 0.432 70 | 0.215 71 | 0.493 84 | 0.599 77 | 0.638 69 | 0.617 95 | 0.570 62 | 0.897 43 | 0.806 56 | 0.605 63 | |
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 | ||||||||||||||||||||||
VACNN++ | 0.684 45 | 0.728 72 | 0.757 52 | 0.776 49 | 0.690 35 | 0.804 64 | 0.464 53 | 0.816 36 | 0.577 57 | 0.587 48 | 0.945 60 | 0.508 41 | 0.276 43 | 0.671 34 | 0.710 60 | 0.663 57 | 0.750 55 | 0.589 57 | 0.881 54 | 0.832 42 | 0.653 45 | |
PointTransformer++ | 0.725 30 | 0.727 73 | 0.811 18 | 0.819 26 | 0.765 11 | 0.841 27 | 0.502 33 | 0.814 39 | 0.621 34 | 0.623 31 | 0.955 27 | 0.556 23 | 0.284 38 | 0.620 49 | 0.866 21 | 0.781 10 | 0.757 51 | 0.648 28 | 0.932 21 | 0.862 22 | 0.709 28 | |
LAP-D | 0.594 84 | 0.720 74 | 0.692 82 | 0.637 91 | 0.456 95 | 0.773 80 | 0.391 86 | 0.730 63 | 0.587 51 | 0.445 87 | 0.940 74 | 0.381 88 | 0.288 35 | 0.434 92 | 0.453 93 | 0.591 83 | 0.649 88 | 0.581 60 | 0.777 89 | 0.749 89 | 0.610 59 | |
DPC | 0.592 85 | 0.720 74 | 0.700 76 | 0.602 95 | 0.480 91 | 0.762 85 | 0.380 89 | 0.713 69 | 0.585 54 | 0.437 88 | 0.940 74 | 0.369 90 | 0.288 35 | 0.434 92 | 0.509 88 | 0.590 85 | 0.639 93 | 0.567 66 | 0.772 91 | 0.755 87 | 0.592 69 | |
Francis Engelmann, Theodora Kontogianni, Bastian Leibe: Dilated Point Convolutions: On the Receptive Field Size of Point Convolutions on 3D Point Clouds. ICRA 2020 | ||||||||||||||||||||||
PointMRNet | 0.640 64 | 0.717 76 | 0.701 75 | 0.692 70 | 0.576 73 | 0.801 65 | 0.467 52 | 0.716 66 | 0.563 64 | 0.459 83 | 0.953 36 | 0.429 73 | 0.169 90 | 0.581 57 | 0.854 30 | 0.605 77 | 0.710 66 | 0.550 75 | 0.894 46 | 0.793 67 | 0.575 72 | |
DGNet | 0.684 45 | 0.712 77 | 0.784 36 | 0.782 48 | 0.658 44 | 0.835 34 | 0.499 38 | 0.823 35 | 0.641 26 | 0.597 44 | 0.950 46 | 0.487 47 | 0.281 40 | 0.575 59 | 0.619 75 | 0.647 65 | 0.764 43 | 0.620 42 | 0.871 69 | 0.846 37 | 0.688 35 | |
FusionNet | 0.688 43 | 0.704 78 | 0.741 64 | 0.754 56 | 0.656 45 | 0.829 39 | 0.501 34 | 0.741 60 | 0.609 40 | 0.548 55 | 0.950 46 | 0.522 36 | 0.371 4 | 0.633 44 | 0.756 50 | 0.715 35 | 0.771 39 | 0.623 40 | 0.861 74 | 0.814 52 | 0.658 43 | |
Feihu Zhang, Jin Fang, Benjamin Wah, Philip Torr: Deep FusionNet for Point Cloud Semantic Segmentation. ECCV 2020 | ||||||||||||||||||||||
PointASNL | 0.666 54 | 0.703 79 | 0.781 38 | 0.751 58 | 0.655 46 | 0.830 38 | 0.471 48 | 0.769 50 | 0.474 87 | 0.537 59 | 0.951 42 | 0.475 52 | 0.279 42 | 0.635 42 | 0.698 65 | 0.675 52 | 0.751 53 | 0.553 73 | 0.816 85 | 0.806 56 | 0.703 31 | |
Xu Yan, Chaoda Zheng, Zhen Li, Sheng Wang, Shuguang Cui: PointASNL: Robust Point Clouds Processing using Nonlocal Neural Networks with Adaptive Sampling. CVPR 2020 | ||||||||||||||||||||||
dtc_net | 0.625 78 | 0.703 79 | 0.751 55 | 0.794 42 | 0.535 82 | 0.848 21 | 0.480 45 | 0.676 77 | 0.528 71 | 0.469 80 | 0.944 66 | 0.454 61 | 0.004 111 | 0.464 88 | 0.636 73 | 0.704 40 | 0.758 49 | 0.548 77 | 0.924 25 | 0.787 74 | 0.492 94 | |
HPGCNN | 0.656 59 | 0.698 81 | 0.743 62 | 0.650 84 | 0.564 76 | 0.820 50 | 0.505 32 | 0.758 53 | 0.631 30 | 0.479 77 | 0.945 60 | 0.480 50 | 0.226 64 | 0.572 60 | 0.774 49 | 0.690 47 | 0.735 59 | 0.614 43 | 0.853 77 | 0.776 80 | 0.597 67 | |
Jisheng Dang, Qingyong Hu, Yulan Guo, Jun Yang: HPGCNN. | ||||||||||||||||||||||
FCPN | 0.447 102 | 0.679 82 | 0.604 104 | 0.578 98 | 0.380 101 | 0.682 97 | 0.291 101 | 0.106 111 | 0.483 85 | 0.258 109 | 0.920 93 | 0.258 105 | 0.025 109 | 0.231 108 | 0.325 100 | 0.480 97 | 0.560 101 | 0.463 97 | 0.725 96 | 0.666 102 | 0.231 111 | |
Dario Rethage, Johanna Wald, Jürgen Sturm, Nassir Navab, Federico Tombari: Fully-Convolutional Point Networks for Large-Scale Point Clouds. ECCV 2018 | ||||||||||||||||||||||
subcloud_weak | 0.516 95 | 0.676 83 | 0.591 105 | 0.609 92 | 0.442 96 | 0.774 79 | 0.335 94 | 0.597 87 | 0.422 99 | 0.357 99 | 0.932 86 | 0.341 95 | 0.094 104 | 0.298 101 | 0.528 86 | 0.473 98 | 0.676 81 | 0.495 91 | 0.602 106 | 0.721 95 | 0.349 107 | |
SQN_0.1% | 0.569 89 | 0.676 83 | 0.696 79 | 0.657 81 | 0.497 87 | 0.779 78 | 0.424 73 | 0.548 95 | 0.515 75 | 0.376 95 | 0.902 101 | 0.422 78 | 0.357 8 | 0.379 97 | 0.456 92 | 0.596 82 | 0.659 86 | 0.544 78 | 0.685 100 | 0.665 103 | 0.556 82 | |
TextureNet | 0.566 90 | 0.672 85 | 0.664 90 | 0.671 77 | 0.494 89 | 0.719 92 | 0.445 64 | 0.678 76 | 0.411 100 | 0.396 93 | 0.935 80 | 0.356 92 | 0.225 66 | 0.412 94 | 0.535 83 | 0.565 89 | 0.636 94 | 0.464 96 | 0.794 88 | 0.680 100 | 0.568 76 | |
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 | ||||||||||||||||||||||
Supervoxel-CNN | 0.635 69 | 0.656 86 | 0.711 70 | 0.719 62 | 0.613 60 | 0.757 86 | 0.444 67 | 0.765 51 | 0.534 69 | 0.566 51 | 0.928 88 | 0.478 51 | 0.272 46 | 0.636 41 | 0.531 84 | 0.664 56 | 0.645 90 | 0.508 88 | 0.864 73 | 0.792 70 | 0.611 57 | |
DVVNet | 0.562 91 | 0.648 87 | 0.700 76 | 0.770 50 | 0.586 70 | 0.687 96 | 0.333 95 | 0.650 80 | 0.514 76 | 0.475 79 | 0.906 98 | 0.359 91 | 0.223 68 | 0.340 99 | 0.442 94 | 0.422 102 | 0.668 84 | 0.501 89 | 0.708 98 | 0.779 77 | 0.534 87 | |
SparseConvNet | 0.725 30 | 0.647 88 | 0.821 9 | 0.846 12 | 0.721 30 | 0.869 5 | 0.533 21 | 0.754 55 | 0.603 44 | 0.614 33 | 0.955 27 | 0.572 17 | 0.325 20 | 0.710 31 | 0.870 19 | 0.724 29 | 0.823 2 | 0.628 37 | 0.934 18 | 0.865 21 | 0.683 36 | |
wsss-transformer | 0.600 83 | 0.634 89 | 0.743 62 | 0.697 69 | 0.601 64 | 0.781 75 | 0.437 71 | 0.585 90 | 0.493 81 | 0.446 85 | 0.933 85 | 0.394 85 | 0.011 110 | 0.654 37 | 0.661 72 | 0.603 78 | 0.733 60 | 0.526 85 | 0.832 81 | 0.761 85 | 0.480 97 | |
3DSM_DMMF | 0.631 72 | 0.626 90 | 0.745 60 | 0.801 39 | 0.607 61 | 0.751 87 | 0.506 31 | 0.729 64 | 0.565 62 | 0.491 75 | 0.866 105 | 0.434 69 | 0.197 81 | 0.595 52 | 0.630 74 | 0.709 37 | 0.705 71 | 0.560 68 | 0.875 60 | 0.740 90 | 0.491 95 | |
MSP | 0.748 18 | 0.623 91 | 0.804 22 | 0.859 3 | 0.745 23 | 0.824 45 | 0.501 34 | 0.912 3 | 0.690 9 | 0.685 7 | 0.956 24 | 0.567 18 | 0.320 22 | 0.768 12 | 0.918 4 | 0.720 31 | 0.802 14 | 0.676 19 | 0.921 27 | 0.881 9 | 0.779 5 | |
EQ-Net | 0.743 23 | 0.620 92 | 0.799 27 | 0.849 9 | 0.730 27 | 0.822 47 | 0.493 41 | 0.897 9 | 0.664 17 | 0.681 8 | 0.955 27 | 0.562 21 | 0.378 3 | 0.760 16 | 0.903 8 | 0.738 23 | 0.801 18 | 0.673 23 | 0.907 34 | 0.877 12 | 0.745 11 | |
Zetong Yang*, Li Jiang*, Yanan Sun, Bernt Schiele, Jiaya JIa: A Unified Query-based Paradigm for Point Cloud Understanding. CVPR 2022 | ||||||||||||||||||||||
joint point-based | 0.634 70 | 0.614 93 | 0.778 39 | 0.667 79 | 0.633 57 | 0.825 43 | 0.420 75 | 0.804 41 | 0.467 89 | 0.561 52 | 0.951 42 | 0.494 43 | 0.291 34 | 0.566 62 | 0.458 91 | 0.579 87 | 0.764 43 | 0.559 70 | 0.838 80 | 0.814 52 | 0.598 66 | |
Hung-Yueh Chiang, Yen-Liang Lin, Yueh-Cheng Liu, Winston H. Hsu: A Unified Point-Based Framework for 3D Segmentation. 3DV 2019 | ||||||||||||||||||||||
Online SegFusion | 0.515 96 | 0.607 94 | 0.644 96 | 0.579 97 | 0.434 97 | 0.630 104 | 0.353 92 | 0.628 85 | 0.440 95 | 0.410 91 | 0.762 110 | 0.307 98 | 0.167 91 | 0.520 75 | 0.403 97 | 0.516 92 | 0.565 98 | 0.447 100 | 0.678 101 | 0.701 97 | 0.514 92 | |
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 | ||||||||||||||||||||||
FusionAwareConv | 0.630 75 | 0.604 95 | 0.741 64 | 0.766 53 | 0.590 67 | 0.747 88 | 0.501 34 | 0.734 62 | 0.503 79 | 0.527 63 | 0.919 94 | 0.454 61 | 0.323 21 | 0.550 70 | 0.420 95 | 0.678 51 | 0.688 77 | 0.544 78 | 0.896 44 | 0.795 64 | 0.627 55 | |
Jiazhao Zhang, Chenyang Zhu, Lintao Zheng, Kai Xu: Fusion-Aware Point Convolution for Online Semantic 3D Scene Segmentation. CVPR 2020 | ||||||||||||||||||||||
PointNet++ | 0.339 110 | 0.584 96 | 0.478 110 | 0.458 106 | 0.256 111 | 0.360 112 | 0.250 103 | 0.247 109 | 0.278 110 | 0.261 108 | 0.677 111 | 0.183 110 | 0.117 101 | 0.212 110 | 0.145 111 | 0.364 105 | 0.346 112 | 0.232 112 | 0.548 108 | 0.523 111 | 0.252 110 | |
Charles R. Qi, Li Yi, Hao Su, Leonidas J. Guibas: pointnet++: deep hierarchical feature learning on point sets in a metric space. | ||||||||||||||||||||||
PointCNN with RGB | 0.458 101 | 0.577 97 | 0.611 101 | 0.356 111 | 0.321 108 | 0.715 93 | 0.299 100 | 0.376 105 | 0.328 107 | 0.319 102 | 0.944 66 | 0.285 101 | 0.164 92 | 0.216 109 | 0.229 104 | 0.484 96 | 0.545 103 | 0.456 98 | 0.755 92 | 0.709 96 | 0.475 99 | |
Yangyan Li, Rui Bu, Mingchao Sun, Baoquan Chen: PointCNN. NeurIPS 2018 | ||||||||||||||||||||||
PCNN | 0.498 98 | 0.559 98 | 0.644 96 | 0.560 99 | 0.420 99 | 0.711 94 | 0.229 106 | 0.414 101 | 0.436 96 | 0.352 100 | 0.941 72 | 0.324 97 | 0.155 95 | 0.238 106 | 0.387 98 | 0.493 94 | 0.529 105 | 0.509 86 | 0.813 86 | 0.751 88 | 0.504 93 | |
3DMV, FTSDF | 0.501 97 | 0.558 99 | 0.608 103 | 0.424 109 | 0.478 92 | 0.690 95 | 0.246 104 | 0.586 89 | 0.468 88 | 0.450 84 | 0.911 96 | 0.394 85 | 0.160 94 | 0.438 90 | 0.212 106 | 0.432 101 | 0.541 104 | 0.475 95 | 0.742 94 | 0.727 93 | 0.477 98 | |
MVF-GNN | 0.658 57 | 0.558 99 | 0.751 55 | 0.655 82 | 0.690 35 | 0.722 91 | 0.453 56 | 0.867 15 | 0.579 56 | 0.576 49 | 0.893 102 | 0.523 35 | 0.293 32 | 0.733 27 | 0.571 80 | 0.692 44 | 0.659 86 | 0.606 47 | 0.875 60 | 0.804 58 | 0.668 40 | |
PNET2 | 0.442 104 | 0.548 101 | 0.548 106 | 0.597 96 | 0.363 104 | 0.628 105 | 0.300 98 | 0.292 106 | 0.374 102 | 0.307 103 | 0.881 103 | 0.268 104 | 0.186 84 | 0.238 106 | 0.204 108 | 0.407 103 | 0.506 109 | 0.449 99 | 0.667 102 | 0.620 106 | 0.462 101 | |
3DWSSS | 0.425 107 | 0.525 102 | 0.647 94 | 0.522 100 | 0.324 107 | 0.488 111 | 0.077 112 | 0.712 70 | 0.353 104 | 0.401 92 | 0.636 112 | 0.281 102 | 0.176 87 | 0.340 99 | 0.565 81 | 0.175 112 | 0.551 102 | 0.398 106 | 0.370 112 | 0.602 107 | 0.361 105 | |
SurfaceConvPF | 0.442 104 | 0.505 103 | 0.622 100 | 0.380 110 | 0.342 106 | 0.654 99 | 0.227 107 | 0.397 103 | 0.367 103 | 0.276 106 | 0.924 90 | 0.240 106 | 0.198 80 | 0.359 98 | 0.262 102 | 0.366 104 | 0.581 96 | 0.435 103 | 0.640 103 | 0.668 101 | 0.398 102 | |
Hao Pan, Shilin Liu, Yang Liu, Xin Tong: Convolutional Neural Networks on 3D Surfaces Using Parallel Frames. | ||||||||||||||||||||||
GMLPs | 0.538 93 | 0.495 104 | 0.693 81 | 0.647 86 | 0.471 93 | 0.793 70 | 0.300 98 | 0.477 100 | 0.505 78 | 0.358 98 | 0.903 100 | 0.327 96 | 0.081 105 | 0.472 87 | 0.529 85 | 0.448 100 | 0.710 66 | 0.509 86 | 0.746 93 | 0.737 91 | 0.554 83 | |
PanopticFusion-label | 0.529 94 | 0.491 105 | 0.688 85 | 0.604 94 | 0.386 100 | 0.632 103 | 0.225 108 | 0.705 72 | 0.434 97 | 0.293 104 | 0.815 106 | 0.348 94 | 0.241 62 | 0.499 82 | 0.669 68 | 0.507 93 | 0.649 88 | 0.442 102 | 0.796 87 | 0.602 107 | 0.561 79 | |
Gaku Narita, Takashi Seno, Tomoya Ishikawa, Yohsuke Kaji: PanopticFusion: Online Volumetric Semantic Mapping at the Level of Stuff and Things. IROS 2019 (to appear) | ||||||||||||||||||||||
3DMV | 0.484 100 | 0.484 106 | 0.538 107 | 0.643 89 | 0.424 98 | 0.606 107 | 0.310 96 | 0.574 92 | 0.433 98 | 0.378 94 | 0.796 108 | 0.301 99 | 0.214 72 | 0.537 73 | 0.208 107 | 0.472 99 | 0.507 108 | 0.413 105 | 0.693 99 | 0.602 107 | 0.539 85 | |
Angela Dai, Matthias Niessner: 3DMV: Joint 3D-Multi-View Prediction for 3D Semantic Scene Segmentation. ECCV'18 | ||||||||||||||||||||||
DGCNN_reproduce | 0.446 103 | 0.474 107 | 0.623 99 | 0.463 105 | 0.366 103 | 0.651 100 | 0.310 96 | 0.389 104 | 0.349 105 | 0.330 101 | 0.937 77 | 0.271 103 | 0.126 100 | 0.285 102 | 0.224 105 | 0.350 108 | 0.577 97 | 0.445 101 | 0.625 104 | 0.723 94 | 0.394 103 | |
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 | ||||||||||||||||||||||
SPLAT Net | 0.393 108 | 0.472 108 | 0.511 108 | 0.606 93 | 0.311 109 | 0.656 98 | 0.245 105 | 0.405 102 | 0.328 107 | 0.197 110 | 0.927 89 | 0.227 108 | 0.000 113 | 0.001 113 | 0.249 103 | 0.271 111 | 0.510 106 | 0.383 108 | 0.593 107 | 0.699 98 | 0.267 109 | |
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 | ||||||||||||||||||||||
Tangent Convolutions | 0.438 106 | 0.437 109 | 0.646 95 | 0.474 104 | 0.369 102 | 0.645 101 | 0.353 92 | 0.258 108 | 0.282 109 | 0.279 105 | 0.918 95 | 0.298 100 | 0.147 99 | 0.283 103 | 0.294 101 | 0.487 95 | 0.562 100 | 0.427 104 | 0.619 105 | 0.633 105 | 0.352 106 | |
Maxim Tatarchenko, Jaesik Park, Vladlen Koltun, Qian-Yi Zhou: Tangent convolutions for dense prediction in 3d. CVPR 2018 | ||||||||||||||||||||||
SSC-UNet | 0.308 111 | 0.353 110 | 0.290 112 | 0.278 112 | 0.166 112 | 0.553 109 | 0.169 110 | 0.286 107 | 0.147 112 | 0.148 112 | 0.908 97 | 0.182 111 | 0.064 107 | 0.023 112 | 0.018 113 | 0.354 107 | 0.363 110 | 0.345 110 | 0.546 110 | 0.685 99 | 0.278 108 | |
ScanNet+FTSDF | 0.383 109 | 0.297 111 | 0.491 109 | 0.432 108 | 0.358 105 | 0.612 106 | 0.274 102 | 0.116 110 | 0.411 100 | 0.265 107 | 0.904 99 | 0.229 107 | 0.079 106 | 0.250 104 | 0.185 109 | 0.320 109 | 0.510 106 | 0.385 107 | 0.548 108 | 0.597 110 | 0.394 103 | |
ScanNet | 0.306 112 | 0.203 112 | 0.366 111 | 0.501 101 | 0.311 109 | 0.524 110 | 0.211 109 | 0.002 113 | 0.342 106 | 0.189 111 | 0.786 109 | 0.145 112 | 0.102 102 | 0.245 105 | 0.152 110 | 0.318 110 | 0.348 111 | 0.300 111 | 0.460 111 | 0.437 112 | 0.182 112 | |
Angela Dai, Angel X. Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, Matthias Nießner: ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes. CVPR'17 | ||||||||||||||||||||||
ERROR | 0.054 113 | 0.000 113 | 0.041 113 | 0.172 113 | 0.030 113 | 0.062 113 | 0.001 113 | 0.035 112 | 0.004 113 | 0.051 113 | 0.143 113 | 0.019 113 | 0.003 112 | 0.041 111 | 0.050 112 | 0.003 113 | 0.054 113 | 0.018 113 | 0.005 113 | 0.264 113 | 0.082 113 | |