3D Semantic Instance Results - ScanNet Benchmark

The 3D semantic instance prediction task involves detecting and segmenting the object in an 3D scan mesh.

Evaluation and metrics

Our evaluation ranks all methods according to the average precision for each class. We report the mean average precision AP at overlap 0.25 (AP 25%), overlap 0.5 (AP 50%), and over overlaps in the range [0.5:0.95:0.05] (AP). Note that multiple predictions of the same ground truth instance are penalized as false positives.

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

Method	avg ap 25%	bathtub	bed	bookshelf	cabinet	chair	counter	curtain	desk	door	otherfurniture	picture	refrigerator	shower curtain	sink	sofa	table	toilet	window

Volt-SPFormerScanNet	0.908 1	1.000 1	0.981 22	0.975 1	0.885 1	0.964 6	0.744 23	0.845 18	0.906 16	0.916 2	0.842 1	0.820 2	0.879 1	0.959 52	0.955 9	0.944 1	0.872 18	0.999 41	0.869 3
Kadir Yilmaz, Adrian Kruse, Tristan Höfer, Daan de Geus, Bastian Leibe: Volume Transformer: Revisiting Vanilla Transformers for 3D Scene Understanding.
SoftGroup++	0.874 17	1.000 1	0.972 29	0.947 2	0.839 15	0.898 31	0.556 45	0.913 2	0.881 25	0.756 29	0.828 3	0.748 24	0.821 2	1.000 1	0.937 15	0.937 2	0.887 5	1.000 1	0.821 13

PointRel	0.901 2	1.000 1	0.978 26	0.928 4	0.879 2	0.962 8	0.882 5	0.749 41	0.947 3	0.912 3	0.802 4	0.753 22	0.820 3	1.000 1	0.984 4	0.919 7	0.894 3	1.000 1	0.815 17
: Relation3D: Enhancing Relation Modeling for Point Cloud Instance Segmentation. CVPR 2025
Mask3D	0.870 19	1.000 1	0.985 20	0.782 50	0.818 24	0.938 21	0.760 20	0.749 41	0.923 11	0.877 17	0.760 14	0.785 16	0.820 3	1.000 1	0.912 23	0.864 40	0.878 12	0.983 57	0.825 11
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
EV3D	0.877 14	1.000 1	0.996 10	0.873 16	0.854 9	0.950 15	0.691 30	0.783 32	0.926 9	0.889 15	0.754 18	0.794 14	0.820 3	1.000 1	0.912 23	0.900 12	0.860 23	1.000 1	0.779 24

Spherical Mask(CtoF)	0.875 15	1.000 1	0.991 17	0.873 16	0.850 10	0.946 17	0.691 30	0.752 40	0.926 9	0.889 14	0.759 15	0.794 13	0.820 3	1.000 1	0.912 23	0.900 12	0.878 12	1.000 1	0.769 26

Queryformer	0.874 17	1.000 1	0.978 25	0.809 40	0.876 3	0.936 22	0.702 27	0.716 46	0.920 13	0.875 18	0.766 13	0.772 18	0.818 7	1.000 1	0.995 1	0.916 8	0.892 4	1.000 1	0.767 27

IPCA-Inst	0.851 24	1.000 1	0.968 31	0.884 13	0.842 14	0.862 44	0.693 29	0.812 26	0.888 24	0.677 41	0.783 9	0.698 30	0.807 8	1.000 1	0.911 31	0.865 39	0.865 22	1.000 1	0.757 30

SoftGroup	0.865 22	1.000 1	0.969 30	0.860 22	0.860 6	0.913 27	0.558 42	0.899 4	0.911 15	0.760 28	0.828 2	0.736 26	0.802 9	0.981 48	0.919 20	0.875 27	0.877 14	1.000 1	0.820 15
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral]
SIM3D	0.878 13	1.000 1	0.972 28	0.863 21	0.817 25	0.952 12	0.821 11	0.783 31	0.890 22	0.902 9	0.735 24	0.797 10	0.799 10	1.000 1	0.931 18	0.893 16	0.853 25	1.000 1	0.792 21

ExtMask3D	0.867 21	1.000 1	1.000 1	0.756 58	0.816 26	0.940 19	0.795 14	0.760 38	0.862 27	0.888 16	0.739 22	0.763 20	0.774 11	1.000 1	0.929 19	0.878 26	0.879 9	1.000 1	0.819 16

ODIN - Ins	0.847 26	1.000 1	0.951 37	0.834 33	0.828 19	0.875 36	0.871 6	0.767 36	0.821 37	0.816 24	0.690 35	0.800 9	0.771 12	1.000 1	0.912 23	0.891 17	0.821 31	0.886 73	0.713 37
Ayush Jain, Pushkal Katara, Nikolaos Gkanatsios, Adam W. Harley, Gabriel Sarch, Kriti Aggarwal, Vishrav Chaudhary, Katerina Fragkiadaki: ODIN: A Single Model for 2D and 3D Segmentation. CVPR 2024
Competitor-MAFT	0.896 4	1.000 1	1.000 1	0.872 18	0.847 12	0.967 5	0.955 1	0.778 35	0.901 19	0.919 1	0.784 8	0.812 4	0.770 13	1.000 1	0.949 10	0.865 38	0.868 20	1.000 1	0.840 7

PointComp	0.897 3	1.000 1	0.998 6	0.864 20	0.869 4	0.969 4	0.830 8	0.783 33	0.905 17	0.894 11	0.791 5	0.834 1	0.769 14	1.000 1	0.982 5	0.920 6	0.868 21	1.000 1	0.872 2

TST3D	0.879 12	1.000 1	0.994 11	0.921 6	0.807 27	0.939 20	0.771 17	0.887 7	0.923 12	0.862 19	0.722 26	0.768 19	0.756 15	1.000 1	0.910 34	0.904 9	0.836 30	0.999 41	0.824 12
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
OneFormer3D	0.896 4	1.000 1	1.000 1	0.913 7	0.858 7	0.951 14	0.786 16	0.837 20	0.916 14	0.908 5	0.778 11	0.803 8	0.750 16	1.000 1	0.976 7	0.926 5	0.882 7	0.995 51	0.849 4
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
MAFT	0.860 23	1.000 1	0.990 18	0.810 39	0.829 18	0.949 16	0.809 12	0.688 52	0.836 32	0.904 7	0.751 19	0.796 11	0.741 17	1.000 1	0.864 44	0.848 49	0.837 28	1.000 1	0.828 9

MG-Former	0.887 6	1.000 1	0.991 16	0.837 28	0.801 28	0.935 23	0.887 4	0.857 11	0.946 4	0.891 12	0.748 21	0.805 7	0.739 18	1.000 1	0.993 2	0.809 62	0.876 15	1.000 1	0.842 6

TopoSeg	0.832 30	1.000 1	0.981 23	0.933 3	0.819 23	0.826 53	0.524 51	0.841 19	0.811 38	0.681 40	0.759 16	0.687 31	0.727 19	0.981 48	0.911 31	0.883 22	0.853 26	1.000 1	0.756 31

KmaxOneFormerNet	0.883 8	1.000 1	1.000 1	0.798 43	0.848 11	0.971 2	0.853 7	0.903 3	0.827 35	0.910 4	0.748 20	0.809 6	0.724 20	1.000 1	0.980 6	0.855 44	0.844 27	1.000 1	0.832 8

UniPerception	0.870 19	1.000 1	0.998 6	0.770 53	0.835 16	0.972 1	0.762 19	0.754 39	0.928 8	0.845 21	0.790 6	0.819 3	0.717 21	0.981 48	0.915 22	0.890 18	0.878 10	1.000 1	0.809 18

TD3D	0.875 15	1.000 1	0.976 27	0.877 14	0.783 34	0.970 3	0.889 3	0.828 21	0.945 5	0.803 27	0.713 28	0.720 29	0.709 22	1.000 1	0.936 16	0.934 4	0.873 16	1.000 1	0.791 22
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
DCD	0.885 7	1.000 1	0.933 44	0.856 24	0.832 17	0.959 10	0.930 2	0.858 10	0.802 41	0.859 20	0.767 12	0.796 12	0.709 23	1.000 1	0.971 8	0.871 32	0.904 1	1.000 1	0.874 1

Mask3D_evaluation	0.843 27	1.000 1	0.955 36	0.847 25	0.795 30	0.932 24	0.750 22	0.780 34	0.891 21	0.818 23	0.737 23	0.633 39	0.703 24	1.000 1	0.902 36	0.870 33	0.820 32	0.941 65	0.805 19

RPGN	0.806 35	1.000 1	0.992 14	0.789 45	0.723 53	0.891 32	0.650 33	0.810 27	0.832 33	0.665 44	0.699 32	0.658 33	0.700 25	1.000 1	0.881 38	0.832 53	0.774 35	0.997 44	0.613 54
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
Box2Mask	0.803 36	1.000 1	0.962 35	0.874 15	0.707 57	0.887 35	0.686 32	0.598 60	0.961 1	0.715 35	0.694 33	0.469 57	0.700 25	1.000 1	0.912 23	0.902 11	0.753 42	0.997 44	0.637 48
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022
CSC-Pretrained	0.791 39	1.000 1	0.996 8	0.829 34	0.767 38	0.889 34	0.600 37	0.819 24	0.770 48	0.594 55	0.620 48	0.541 49	0.700 25	1.000 1	0.941 11	0.889 20	0.763 38	1.000 1	0.526 64

InsSSM	0.883 8	1.000 1	0.996 8	0.800 42	0.865 5	0.960 9	0.808 13	0.852 16	0.940 7	0.899 10	0.785 7	0.810 5	0.700 25	1.000 1	0.912 23	0.851 47	0.895 2	0.997 44	0.827 10
Lei Yao, Yi Wang, Moyun Liu, Lap-Pui Chau: SGIFormer: Semantic-guided and Geometric-enhanced Interleaving Transformer for 3D Instance Segmentation. TCSVT, 2024
GICN	0.788 41	1.000 1	0.978 24	0.867 19	0.781 35	0.833 50	0.527 50	0.824 22	0.806 39	0.549 63	0.596 51	0.551 45	0.700 25	1.000 1	0.853 45	0.935 3	0.733 46	1.000 1	0.651 45

SPFormer	0.851 24	1.000 1	0.994 12	0.806 41	0.774 36	0.942 18	0.637 34	0.849 17	0.859 29	0.889 13	0.720 27	0.730 27	0.665 30	1.000 1	0.911 31	0.868 37	0.873 17	1.000 1	0.796 20
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
Competitor-SPFormer	0.881 10	1.000 1	1.000 1	0.845 26	0.854 8	0.962 7	0.714 26	0.857 12	0.904 18	0.902 8	0.782 10	0.789 15	0.662 31	1.000 1	0.988 3	0.874 29	0.886 6	0.997 44	0.847 5

SSEC	0.820 33	1.000 1	0.983 21	0.924 5	0.826 20	0.817 56	0.415 60	0.899 5	0.793 43	0.673 42	0.731 25	0.636 37	0.653 32	1.000 1	0.939 13	0.804 64	0.878 11	1.000 1	0.780 23

SphereSeg	0.835 28	1.000 1	0.963 34	0.891 10	0.794 31	0.954 11	0.822 10	0.710 47	0.961 2	0.721 33	0.693 34	0.530 52	0.653 33	1.000 1	0.867 43	0.857 43	0.859 24	0.991 54	0.771 25

DualGroup	0.782 44	1.000 1	0.927 47	0.811 37	0.772 37	0.853 46	0.631 36	0.805 28	0.773 45	0.613 51	0.611 49	0.610 40	0.650 34	0.835 66	0.881 38	0.879 25	0.750 44	1.000 1	0.675 43

DD-UNet+Group	0.764 48	1.000 1	0.897 60	0.837 29	0.753 44	0.830 52	0.459 56	0.824 22	0.699 57	0.629 48	0.653 39	0.438 60	0.650 34	1.000 1	0.880 40	0.858 42	0.690 59	1.000 1	0.650 46
H. Liu, R. Liu, K. Yang, J. Zhang, K. Peng, R. Stiefelhagen: HIDA: Towards Holistic Indoor Understanding for the Visually Impaired via Semantic Instance Segmentation with a Wearable Solid-State LiDAR Sensor. ICCVW 2021
INS-Conv-instance	0.762 49	1.000 1	0.923 50	0.765 54	0.785 33	0.905 29	0.600 37	0.655 54	0.646 62	0.683 39	0.647 42	0.530 51	0.650 34	1.000 1	0.824 50	0.830 54	0.693 57	0.944 61	0.644 47

DKNet	0.815 34	1.000 1	0.930 45	0.844 27	0.765 40	0.915 26	0.534 49	0.805 28	0.805 40	0.807 26	0.654 38	0.763 21	0.650 34	1.000 1	0.794 57	0.881 23	0.766 37	1.000 1	0.758 29
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
3D-MPA	0.737 53	1.000 1	0.933 43	0.785 46	0.794 32	0.831 51	0.279 68	0.588 61	0.695 58	0.616 49	0.559 58	0.556 44	0.650 34	1.000 1	0.809 54	0.875 28	0.696 55	1.000 1	0.608 56
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020
DENet	0.786 42	1.000 1	0.929 46	0.736 59	0.750 47	0.720 69	0.755 21	0.934 1	0.794 42	0.590 56	0.561 57	0.537 50	0.650 34	1.000 1	0.882 37	0.804 65	0.789 34	1.000 1	0.719 36

PBNet	0.825 32	1.000 1	0.963 33	0.837 30	0.843 13	0.865 39	0.822 9	0.647 55	0.878 26	0.733 31	0.639 44	0.683 32	0.650 34	1.000 1	0.853 45	0.870 34	0.820 33	1.000 1	0.744 32
Weiguang Zhao, Yuyao Yan, Chaolong Yang, Jianan Ye, Xi Yang, Kaizhu Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
PCJC	0.684 64	1.000 1	0.895 61	0.757 57	0.659 62	0.862 43	0.189 75	0.739 44	0.606 65	0.712 36	0.581 54	0.515 54	0.650 34	0.857 58	0.357 76	0.785 68	0.631 64	0.889 71	0.635 49

ISBNet	0.835 28	1.000 1	0.950 38	0.731 60	0.819 22	0.918 25	0.790 15	0.740 43	0.851 31	0.831 22	0.661 37	0.742 25	0.650 34	1.000 1	0.937 14	0.814 61	0.836 29	1.000 1	0.765 28
Tuan Duc Ngo, Binh-Son Hua, Khoi Nguyen: ISBNet: a 3D Point Cloud Instance Segmentation Network with Instance-aware Sampling and Box-aware Dynamic Convolution. CVPR 2023
PE	0.776 46	1.000 1	0.900 58	0.860 22	0.728 52	0.869 37	0.400 61	0.857 13	0.774 44	0.568 62	0.701 31	0.602 41	0.646 43	0.933 54	0.843 48	0.890 19	0.691 58	0.997 44	0.709 38
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
HAIS	0.803 36	1.000 1	0.994 12	0.820 35	0.759 41	0.855 45	0.554 46	0.882 8	0.827 36	0.615 50	0.676 36	0.638 36	0.646 43	1.000 1	0.912 23	0.797 67	0.767 36	0.994 52	0.726 34
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
GraphCut	0.832 30	1.000 1	0.922 53	0.724 62	0.798 29	0.902 30	0.701 28	0.856 14	0.859 28	0.715 34	0.706 29	0.748 23	0.640 45	1.000 1	0.934 17	0.862 41	0.880 8	1.000 1	0.729 33

PointGroup	0.778 45	1.000 1	0.900 57	0.798 44	0.715 55	0.863 41	0.493 52	0.706 48	0.895 20	0.569 61	0.701 30	0.576 43	0.639 46	1.000 1	0.880 40	0.851 46	0.719 50	0.997 44	0.709 39
Li Jiang, Hengshuang Zhao, Shaoshuai Shi, Shu Liu, Chi-Wing Fu, Jiaya Jia: PointGroup: Dual-Set Point Grouping for 3D Instance Segmentation. CVPR 2020 [oral]
ClickSeg_Instance	0.685 63	1.000 1	0.818 69	0.600 71	0.715 56	0.795 58	0.557 43	0.533 66	0.591 68	0.601 53	0.519 64	0.429 62	0.638 47	0.938 53	0.706 65	0.817 59	0.624 66	0.944 61	0.502 66

DANCENET	0.786 42	1.000 1	0.936 41	0.783 48	0.737 50	0.852 47	0.742 24	0.647 55	0.765 50	0.811 25	0.624 47	0.579 42	0.632 48	1.000 1	0.909 35	0.898 14	0.696 54	0.944 61	0.601 57

RWSeg	0.739 52	1.000 1	0.899 59	0.759 56	0.753 45	0.823 54	0.282 66	0.691 51	0.658 60	0.582 59	0.594 52	0.547 46	0.628 49	1.000 1	0.795 56	0.868 36	0.728 48	1.000 1	0.692 41

VDG-Uni3DSeg	0.880 11	1.000 1	0.990 18	0.889 11	0.823 21	0.952 13	0.764 18	0.893 6	0.941 6	0.907 6	0.756 17	0.781 17	0.628 49	1.000 1	0.918 21	0.903 10	0.872 19	0.999 41	0.821 14

Sparse R-CNN	0.714 58	1.000 1	0.926 49	0.694 63	0.699 59	0.890 33	0.636 35	0.516 68	0.693 59	0.743 30	0.588 53	0.369 64	0.601 51	0.594 72	0.800 55	0.886 21	0.676 60	0.986 56	0.546 61

Hier3D	0.540 72	1.000 1	0.727 72	0.626 69	0.467 76	0.693 71	0.200 73	0.412 70	0.480 75	0.528 67	0.318 74	0.077 78	0.600 52	0.688 69	0.382 73	0.768 71	0.472 73	0.941 65	0.350 75
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
SPG_WSIS	0.678 65	1.000 1	0.880 64	0.836 31	0.701 58	0.727 68	0.273 70	0.607 59	0.706 56	0.541 66	0.515 65	0.174 71	0.600 52	0.857 58	0.716 64	0.846 50	0.711 51	1.000 1	0.506 65

SSTNet	0.789 40	1.000 1	0.840 67	0.888 12	0.717 54	0.835 49	0.717 25	0.684 53	0.627 63	0.724 32	0.652 40	0.727 28	0.600 52	1.000 1	0.912 23	0.822 56	0.757 41	1.000 1	0.691 42
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
Mask-Group	0.792 38	1.000 1	0.968 32	0.812 36	0.766 39	0.864 40	0.460 54	0.815 25	0.888 23	0.598 54	0.651 41	0.639 35	0.600 52	0.918 55	0.941 11	0.896 15	0.721 49	1.000 1	0.723 35
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022
3D-SIS	0.558 70	1.000 1	0.773 71	0.614 70	0.503 73	0.691 72	0.200 73	0.412 70	0.498 74	0.546 65	0.311 75	0.103 75	0.600 52	0.857 58	0.382 73	0.799 66	0.445 77	0.938 67	0.371 72
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
AOIA	0.767 47	1.000 1	0.937 40	0.810 38	0.740 49	0.906 28	0.550 47	0.800 30	0.706 55	0.577 60	0.624 46	0.544 48	0.596 57	0.857 58	0.879 42	0.880 24	0.750 43	0.992 53	0.658 44

OccuSeg+instance	0.742 51	1.000 1	0.923 50	0.785 46	0.745 48	0.867 38	0.557 43	0.578 64	0.729 53	0.670 43	0.644 43	0.488 55	0.577 58	1.000 1	0.794 57	0.830 54	0.620 67	1.000 1	0.550 60
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
MTML	0.731 54	1.000 1	0.992 14	0.779 52	0.609 66	0.746 64	0.308 65	0.867 9	0.601 66	0.607 52	0.539 61	0.519 53	0.550 59	1.000 1	0.824 50	0.869 35	0.729 47	1.000 1	0.616 52
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
Dyco3D	0.761 50	1.000 1	0.935 42	0.893 9	0.752 46	0.863 42	0.600 37	0.588 61	0.742 52	0.641 46	0.633 45	0.546 47	0.550 59	0.857 58	0.789 59	0.853 45	0.762 39	0.987 55	0.699 40
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
MASC	0.615 68	0.711 75	0.802 70	0.540 74	0.757 42	0.777 60	0.029 78	0.577 65	0.588 69	0.521 69	0.600 50	0.436 61	0.534 61	0.697 68	0.616 69	0.838 52	0.526 69	0.980 58	0.534 63
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
SegGroup_ins	0.637 67	1.000 1	0.923 52	0.593 72	0.561 70	0.746 65	0.143 77	0.504 69	0.766 49	0.485 71	0.442 69	0.372 63	0.530 62	0.714 67	0.815 53	0.775 69	0.673 61	1.000 1	0.431 70
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
PanopticFusion-inst	0.693 60	1.000 1	0.852 66	0.655 67	0.616 65	0.788 59	0.334 63	0.763 37	0.771 47	0.457 73	0.555 59	0.652 34	0.518 63	0.857 58	0.765 60	0.732 73	0.631 63	0.944 61	0.577 59
Gaku Narita, Takashi Seno, Tomoya Ishikawa, Yohsuke Kaji: PanopticFusion: Online Volumetric Semantic Mapping at the Level of Stuff and Things. IROS 2019 (to appear)
3D-BoNet	0.687 62	1.000 1	0.887 62	0.836 31	0.587 69	0.643 76	0.550 47	0.620 57	0.724 54	0.522 68	0.501 66	0.243 68	0.512 64	1.000 1	0.751 62	0.807 63	0.661 62	0.909 70	0.612 55
Bo Yang, Jianan Wang, Ronald Clark, Qingyong Hu, Sen Wang, Andrew Markham, Niki Trigoni: Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds. NeurIPS 2019 Spotlight
Occipital-SCS	0.688 61	1.000 1	0.913 54	0.730 61	0.737 51	0.743 66	0.442 57	0.855 15	0.655 61	0.546 64	0.546 60	0.263 67	0.508 65	0.889 56	0.568 71	0.771 70	0.705 52	0.889 71	0.625 50

NeuralBF	0.718 57	1.000 1	0.945 39	0.901 8	0.754 43	0.817 55	0.460 54	0.700 49	0.772 46	0.688 38	0.568 56	0.000 79	0.500 66	0.981 48	0.606 70	0.872 30	0.740 45	1.000 1	0.614 53
Weiwei Sun, Daniel Rebain, Renjie Liao, Vladimir Tankovich, Soroosh Yazdani, Kwang Moo Yi, Andrea Tagliasacchi: NeuralBF: Neural Bilateral Filtering for Top-down Instance Segmentation on Point Clouds. WACV 2023
OSIS	0.725 55	1.000 1	0.885 63	0.653 68	0.657 63	0.801 57	0.576 41	0.695 50	0.828 34	0.698 37	0.534 62	0.457 59	0.500 66	0.857 58	0.831 49	0.841 51	0.627 65	1.000 1	0.619 51

SALoss-ResNet	0.695 59	1.000 1	0.855 65	0.579 73	0.589 68	0.735 67	0.484 53	0.588 61	0.856 30	0.634 47	0.571 55	0.298 65	0.500 66	1.000 1	0.824 50	0.818 57	0.702 53	0.935 68	0.545 62
Zhidong Liang, Ming Yang, Hao Li, Chunxiang Wang: 3D Instance Embedding Learning With a Structure-Aware Loss Function for Point Cloud Segmentation. IEEE Robotics and Automation Letters (IROS2020)
SSEN	0.724 56	1.000 1	0.926 48	0.781 51	0.661 61	0.845 48	0.596 40	0.529 67	0.764 51	0.653 45	0.489 68	0.461 58	0.500 66	0.859 57	0.765 60	0.872 31	0.761 40	1.000 1	0.577 58
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv
One_Thing_One_Click	0.675 66	1.000 1	0.823 68	0.782 49	0.621 64	0.766 61	0.211 72	0.736 45	0.560 70	0.586 57	0.522 63	0.636 38	0.453 70	0.641 70	0.853 45	0.850 48	0.694 56	0.997 44	0.411 71
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
ASIS	0.422 77	0.333 79	0.707 75	0.676 65	0.401 77	0.650 75	0.350 62	0.177 78	0.594 67	0.376 76	0.202 77	0.077 77	0.404 71	0.571 73	0.197 80	0.674 77	0.447 76	0.500 79	0.260 79

UNet-backbone	0.605 69	1.000 1	0.909 55	0.764 55	0.603 67	0.704 70	0.415 59	0.301 74	0.548 71	0.461 72	0.394 70	0.267 66	0.386 72	0.857 58	0.649 68	0.817 58	0.504 71	0.959 59	0.356 74

SemRegionNet-20cls	0.470 76	1.000 1	0.727 72	0.447 77	0.481 74	0.678 73	0.024 79	0.380 72	0.518 72	0.440 74	0.339 72	0.128 73	0.350 73	0.429 75	0.212 79	0.711 75	0.465 74	0.833 76	0.290 78

tmp	0.474 75	1.000 1	0.727 72	0.433 78	0.481 75	0.673 74	0.022 80	0.380 72	0.517 73	0.436 75	0.338 73	0.128 73	0.343 74	0.429 75	0.291 78	0.728 74	0.473 72	0.833 76	0.300 77

3D-BEVIS	0.401 78	0.667 76	0.687 76	0.419 79	0.137 80	0.587 78	0.188 76	0.235 75	0.359 77	0.211 79	0.093 80	0.080 76	0.311 75	0.571 73	0.382 73	0.754 72	0.300 79	0.874 74	0.357 73
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
R-PointNet	0.544 71	0.500 78	0.655 77	0.661 66	0.663 60	0.765 62	0.432 58	0.214 77	0.612 64	0.584 58	0.499 67	0.204 70	0.286 76	0.429 75	0.655 67	0.650 78	0.539 68	0.950 60	0.499 67

Sem_Recon_ins	0.484 74	0.764 74	0.608 79	0.470 76	0.521 72	0.637 77	0.311 64	0.218 76	0.348 78	0.365 77	0.223 76	0.222 69	0.258 77	0.629 71	0.734 63	0.596 79	0.509 70	0.858 75	0.444 69

Sgpn_scannet	0.390 79	0.556 77	0.636 78	0.493 75	0.353 78	0.539 79	0.271 71	0.160 79	0.450 76	0.359 78	0.178 78	0.146 72	0.250 78	0.143 79	0.347 77	0.698 76	0.436 78	0.667 78	0.331 76

Region-18class	0.497 73	0.250 80	0.902 56	0.689 64	0.540 71	0.747 63	0.276 69	0.610 58	0.268 79	0.489 70	0.348 71	0.000 79	0.243 79	0.220 78	0.663 66	0.814 60	0.459 75	0.928 69	0.496 68

MaskRCNN 2d->3d Proj	0.261 80	0.903 73	0.081 80	0.008 80	0.233 79	0.175 80	0.280 67	0.106 80	0.150 80	0.203 80	0.175 79	0.480 56	0.218 80	0.143 79	0.542 72	0.404 80	0.153 80	0.393 80	0.049 80

3D Semantic Instance Benchmark