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

UniPerception	0.884 6	1.000 1	0.979 20	0.872 16	0.869 3	0.892 28	0.806 12	0.890 6	0.835 29	0.892 9	0.755 14	0.811 2	0.779 10	0.955 48	0.951 8	0.876 22	0.914 1	0.997 40	0.840 6

DCD	0.885 5	1.000 1	0.933 40	0.856 22	0.832 15	0.959 7	0.930 2	0.858 10	0.802 37	0.859 18	0.767 9	0.796 9	0.709 20	1.000 1	0.971 7	0.871 28	0.904 2	1.000 1	0.874 1

InsSSM	0.883 7	1.000 1	0.996 6	0.800 39	0.865 4	0.960 6	0.808 11	0.852 16	0.940 6	0.899 8	0.785 4	0.810 3	0.700 22	1.000 1	0.912 20	0.851 43	0.895 3	0.997 40	0.827 9
Lei Yao, Yi Wang, Moyun Liu, Lap-Pui Chau: SGIFormer: Semantic-guided and Geometric-enhanced Interleaving Transformer for 3D Instance Segmentation. TCSVT, 2024
PointRel	0.901 1	1.000 1	0.978 23	0.928 3	0.879 1	0.962 5	0.882 5	0.749 37	0.947 3	0.912 2	0.802 3	0.753 18	0.820 2	1.000 1	0.984 4	0.919 5	0.894 4	1.000 1	0.815 15
: Relation3D (PointRel): Enhancing Relation Modeling for Point Cloud Instance Segmentation.
Queryformer	0.874 15	1.000 1	0.978 22	0.809 37	0.876 2	0.936 18	0.702 23	0.716 42	0.920 11	0.875 16	0.766 10	0.772 14	0.818 6	1.000 1	0.995 1	0.916 6	0.892 5	1.000 1	0.767 24

SoftGroup++	0.874 15	1.000 1	0.972 26	0.947 1	0.839 14	0.898 27	0.556 41	0.913 2	0.881 21	0.756 25	0.828 2	0.748 20	0.821 1	1.000 1	0.937 14	0.937 1	0.887 6	1.000 1	0.821 12

Competitor-SPFormer	0.881 9	1.000 1	1.000 1	0.845 24	0.854 7	0.962 4	0.714 22	0.857 12	0.904 14	0.902 6	0.782 7	0.789 12	0.662 28	1.000 1	0.988 3	0.874 25	0.886 7	0.997 40	0.847 3

OneFormer3D	0.896 2	1.000 1	1.000 1	0.913 6	0.858 6	0.951 10	0.786 15	0.837 19	0.916 12	0.908 4	0.778 8	0.803 6	0.750 14	1.000 1	0.976 6	0.926 4	0.882 8	0.995 48	0.849 2
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
GraphCut	0.832 26	1.000 1	0.922 49	0.724 58	0.798 25	0.902 26	0.701 24	0.856 14	0.859 24	0.715 30	0.706 26	0.748 19	0.640 42	1.000 1	0.934 16	0.862 37	0.880 9	1.000 1	0.729 30

ExtMask3D	0.867 18	1.000 1	1.000 1	0.756 54	0.816 22	0.940 15	0.795 13	0.760 35	0.862 23	0.888 14	0.739 19	0.763 16	0.774 11	1.000 1	0.929 18	0.878 21	0.879 10	1.000 1	0.819 14

SSEC	0.820 29	1.000 1	0.983 18	0.924 4	0.826 17	0.817 52	0.415 56	0.899 5	0.793 39	0.673 38	0.731 22	0.636 33	0.653 29	1.000 1	0.939 12	0.804 60	0.878 11	1.000 1	0.780 20

Mask3D	0.870 17	1.000 1	0.985 17	0.782 47	0.818 20	0.938 17	0.760 17	0.749 37	0.923 9	0.877 15	0.760 11	0.785 13	0.820 2	1.000 1	0.912 20	0.864 36	0.878 12	0.983 54	0.825 10
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
Spherical Mask(CtoF)	0.875 13	1.000 1	0.991 15	0.873 14	0.850 9	0.946 13	0.691 26	0.752 36	0.926 7	0.889 12	0.759 12	0.794 10	0.820 2	1.000 1	0.912 20	0.900 9	0.878 12	1.000 1	0.769 23

SoftGroup	0.865 19	1.000 1	0.969 27	0.860 20	0.860 5	0.913 23	0.558 38	0.899 4	0.911 13	0.760 24	0.828 1	0.736 22	0.802 8	0.981 45	0.919 19	0.875 23	0.877 14	1.000 1	0.820 13
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral]
MG-Former	0.887 4	1.000 1	0.991 14	0.837 26	0.801 24	0.935 19	0.887 4	0.857 11	0.946 4	0.891 10	0.748 18	0.805 5	0.739 16	1.000 1	0.993 2	0.809 58	0.876 15	1.000 1	0.842 4

TD3D	0.875 13	1.000 1	0.976 24	0.877 12	0.783 30	0.970 2	0.889 3	0.828 20	0.945 5	0.803 23	0.713 25	0.720 25	0.709 19	1.000 1	0.936 15	0.934 3	0.873 16	1.000 1	0.791 19
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
SPFormer	0.851 21	1.000 1	0.994 10	0.806 38	0.774 32	0.942 14	0.637 30	0.849 17	0.859 25	0.889 11	0.720 24	0.730 23	0.665 27	1.000 1	0.911 27	0.868 33	0.873 17	1.000 1	0.796 17
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
Competitor-MAFT	0.896 2	1.000 1	1.000 1	0.872 16	0.847 11	0.967 3	0.955 1	0.778 33	0.901 15	0.919 1	0.784 5	0.812 1	0.770 12	1.000 1	0.949 9	0.865 34	0.868 18	1.000 1	0.840 5

IPCA-Inst	0.851 21	1.000 1	0.968 28	0.884 11	0.842 13	0.862 40	0.693 25	0.812 25	0.888 20	0.677 37	0.783 6	0.698 26	0.807 7	1.000 1	0.911 27	0.865 35	0.865 19	1.000 1	0.757 27

EV3D	0.877 12	1.000 1	0.996 8	0.873 14	0.854 8	0.950 11	0.691 26	0.783 31	0.926 7	0.889 13	0.754 15	0.794 11	0.820 2	1.000 1	0.912 20	0.900 9	0.860 20	1.000 1	0.779 21

SphereSeg	0.835 24	1.000 1	0.963 31	0.891 9	0.794 27	0.954 8	0.822 8	0.710 43	0.961 2	0.721 29	0.693 31	0.530 48	0.653 30	1.000 1	0.867 39	0.857 39	0.859 21	0.991 51	0.771 22

SIM3D	0.878 11	1.000 1	0.972 25	0.863 19	0.817 21	0.952 9	0.821 9	0.783 30	0.890 18	0.902 7	0.735 21	0.797 7	0.799 9	1.000 1	0.931 17	0.893 13	0.853 22	1.000 1	0.792 18

TopoSeg	0.832 26	1.000 1	0.981 19	0.933 2	0.819 19	0.826 49	0.524 47	0.841 18	0.811 34	0.681 36	0.759 13	0.687 27	0.727 17	0.981 45	0.911 27	0.883 17	0.853 23	1.000 1	0.756 28

KmaxOneFormerNet	0.883 7	1.000 1	1.000 1	0.798 40	0.848 10	0.971 1	0.853 6	0.903 3	0.827 32	0.910 3	0.748 17	0.809 4	0.724 18	1.000 1	0.980 5	0.855 40	0.844 24	1.000 1	0.832 7

MAFT	0.860 20	1.000 1	0.990 16	0.810 36	0.829 16	0.949 12	0.809 10	0.688 48	0.836 28	0.904 5	0.751 16	0.796 8	0.741 15	1.000 1	0.864 40	0.848 45	0.837 25	1.000 1	0.828 8

ISBNet	0.835 24	1.000 1	0.950 34	0.731 56	0.819 18	0.918 21	0.790 14	0.740 39	0.851 27	0.831 19	0.661 33	0.742 21	0.650 31	1.000 1	0.937 13	0.814 57	0.836 26	1.000 1	0.765 25
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
TST3D	0.879 10	1.000 1	0.994 9	0.921 5	0.807 23	0.939 16	0.771 16	0.887 7	0.923 10	0.862 17	0.722 23	0.768 15	0.756 13	1.000 1	0.910 30	0.904 7	0.836 27	0.999 39	0.824 11
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
Mask3D_evaluation	0.843 23	1.000 1	0.955 33	0.847 23	0.795 26	0.932 20	0.750 19	0.780 32	0.891 17	0.818 20	0.737 20	0.633 35	0.703 21	1.000 1	0.902 32	0.870 29	0.820 28	0.941 62	0.805 16

PBNet	0.825 28	1.000 1	0.963 30	0.837 28	0.843 12	0.865 35	0.822 7	0.647 51	0.878 22	0.733 27	0.639 40	0.683 28	0.650 31	1.000 1	0.853 41	0.870 30	0.820 29	1.000 1	0.744 29
Weiguang Zhao, Yuyao Yan, Chaolong Yang, Jianan Ye, Xi Yang, Kaizhu Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
DENet	0.786 38	1.000 1	0.929 42	0.736 55	0.750 43	0.720 65	0.755 18	0.934 1	0.794 38	0.590 52	0.561 53	0.537 46	0.650 31	1.000 1	0.882 33	0.804 61	0.789 30	1.000 1	0.719 33

RPGN	0.806 31	1.000 1	0.992 12	0.789 42	0.723 49	0.891 29	0.650 29	0.810 26	0.832 30	0.665 40	0.699 29	0.658 29	0.700 22	1.000 1	0.881 34	0.832 49	0.774 31	0.997 40	0.613 50
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
HAIS	0.803 32	1.000 1	0.994 10	0.820 32	0.759 37	0.855 41	0.554 42	0.882 8	0.827 33	0.615 46	0.676 32	0.638 32	0.646 40	1.000 1	0.912 20	0.797 63	0.767 32	0.994 49	0.726 31
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
DKNet	0.815 30	1.000 1	0.930 41	0.844 25	0.765 36	0.915 22	0.534 45	0.805 27	0.805 36	0.807 22	0.654 34	0.763 17	0.650 31	1.000 1	0.794 53	0.881 18	0.766 33	1.000 1	0.758 26
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
CSC-Pretrained	0.791 35	1.000 1	0.996 6	0.829 31	0.767 34	0.889 31	0.600 33	0.819 23	0.770 44	0.594 51	0.620 44	0.541 45	0.700 22	1.000 1	0.941 10	0.889 15	0.763 34	1.000 1	0.526 60

Dyco3D	0.761 46	1.000 1	0.935 38	0.893 8	0.752 42	0.863 38	0.600 33	0.588 57	0.742 48	0.641 42	0.633 41	0.546 43	0.550 55	0.857 54	0.789 55	0.853 41	0.762 35	0.987 52	0.699 36
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
SSEN	0.724 52	1.000 1	0.926 44	0.781 48	0.661 57	0.845 44	0.596 36	0.529 63	0.764 47	0.653 41	0.489 64	0.461 54	0.500 62	0.859 53	0.765 56	0.872 27	0.761 36	1.000 1	0.577 54
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv
SSTNet	0.789 36	1.000 1	0.840 63	0.888 10	0.717 50	0.835 45	0.717 21	0.684 49	0.627 59	0.724 28	0.652 36	0.727 24	0.600 48	1.000 1	0.912 20	0.822 52	0.757 37	1.000 1	0.691 38
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
Box2Mask	0.803 32	1.000 1	0.962 32	0.874 13	0.707 53	0.887 32	0.686 28	0.598 56	0.961 1	0.715 31	0.694 30	0.469 53	0.700 22	1.000 1	0.912 20	0.902 8	0.753 38	0.997 40	0.637 44
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022
AOIA	0.767 43	1.000 1	0.937 36	0.810 35	0.740 45	0.906 24	0.550 43	0.800 29	0.706 51	0.577 56	0.624 42	0.544 44	0.596 53	0.857 54	0.879 38	0.880 19	0.750 39	0.992 50	0.658 40

DualGroup	0.782 40	1.000 1	0.927 43	0.811 34	0.772 33	0.853 42	0.631 32	0.805 27	0.773 41	0.613 47	0.611 45	0.610 36	0.650 31	0.835 62	0.881 34	0.879 20	0.750 40	1.000 1	0.675 39

NeuralBF	0.718 53	1.000 1	0.945 35	0.901 7	0.754 39	0.817 51	0.460 50	0.700 45	0.772 42	0.688 34	0.568 52	0.000 75	0.500 62	0.981 45	0.606 66	0.872 26	0.740 41	1.000 1	0.614 49
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
GICN	0.788 37	1.000 1	0.978 21	0.867 18	0.781 31	0.833 46	0.527 46	0.824 21	0.806 35	0.549 59	0.596 47	0.551 41	0.700 22	1.000 1	0.853 41	0.935 2	0.733 42	1.000 1	0.651 41

MTML	0.731 50	1.000 1	0.992 12	0.779 49	0.609 62	0.746 60	0.308 61	0.867 9	0.601 62	0.607 48	0.539 57	0.519 49	0.550 55	1.000 1	0.824 46	0.869 31	0.729 43	1.000 1	0.616 48
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
RWSeg	0.739 48	1.000 1	0.899 55	0.759 52	0.753 41	0.823 50	0.282 62	0.691 47	0.658 56	0.582 55	0.594 48	0.547 42	0.628 46	1.000 1	0.795 52	0.868 32	0.728 44	1.000 1	0.692 37

Mask-Group	0.792 34	1.000 1	0.968 29	0.812 33	0.766 35	0.864 36	0.460 50	0.815 24	0.888 19	0.598 50	0.651 37	0.639 31	0.600 48	0.918 51	0.941 10	0.896 12	0.721 45	1.000 1	0.723 32
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022
PointGroup	0.778 41	1.000 1	0.900 53	0.798 41	0.715 51	0.863 37	0.493 48	0.706 44	0.895 16	0.569 57	0.701 27	0.576 39	0.639 43	1.000 1	0.880 36	0.851 42	0.719 46	0.997 40	0.709 35
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]
SPG_WSIS	0.678 61	1.000 1	0.880 60	0.836 29	0.701 54	0.727 64	0.273 66	0.607 55	0.706 52	0.541 62	0.515 61	0.174 67	0.600 48	0.857 54	0.716 60	0.846 46	0.711 47	1.000 1	0.506 61

Occipital-SCS	0.688 57	1.000 1	0.913 50	0.730 57	0.737 47	0.743 62	0.442 53	0.855 15	0.655 57	0.546 60	0.546 56	0.263 63	0.508 61	0.889 52	0.568 67	0.771 66	0.705 48	0.889 68	0.625 46

SALoss-ResNet	0.695 55	1.000 1	0.855 61	0.579 69	0.589 64	0.735 63	0.484 49	0.588 57	0.856 26	0.634 43	0.571 51	0.298 61	0.500 62	1.000 1	0.824 46	0.818 53	0.702 49	0.935 65	0.545 58
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)
DANCENET	0.786 38	1.000 1	0.936 37	0.783 45	0.737 46	0.852 43	0.742 20	0.647 51	0.765 46	0.811 21	0.624 43	0.579 38	0.632 45	1.000 1	0.909 31	0.898 11	0.696 50	0.944 58	0.601 53

3D-MPA	0.737 49	1.000 1	0.933 39	0.785 43	0.794 28	0.831 47	0.279 64	0.588 57	0.695 54	0.616 45	0.559 54	0.556 40	0.650 31	1.000 1	0.809 50	0.875 24	0.696 51	1.000 1	0.608 52
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020
One_Thing_One_Click	0.675 62	1.000 1	0.823 64	0.782 46	0.621 60	0.766 57	0.211 68	0.736 41	0.560 66	0.586 53	0.522 59	0.636 34	0.453 66	0.641 66	0.853 41	0.850 44	0.694 52	0.997 40	0.411 67
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
INS-Conv-instance	0.762 45	1.000 1	0.923 46	0.765 50	0.785 29	0.905 25	0.600 33	0.655 50	0.646 58	0.683 35	0.647 38	0.530 47	0.650 31	1.000 1	0.824 46	0.830 50	0.693 53	0.944 58	0.644 43

PE	0.776 42	1.000 1	0.900 54	0.860 20	0.728 48	0.869 33	0.400 57	0.857 13	0.774 40	0.568 58	0.701 28	0.602 37	0.646 40	0.933 50	0.843 44	0.890 14	0.691 54	0.997 40	0.709 34
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
DD-UNet+Group	0.764 44	1.000 1	0.897 56	0.837 27	0.753 40	0.830 48	0.459 52	0.824 21	0.699 53	0.629 44	0.653 35	0.438 56	0.650 31	1.000 1	0.880 36	0.858 38	0.690 55	1.000 1	0.650 42
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
Sparse R-CNN	0.714 54	1.000 1	0.926 45	0.694 59	0.699 55	0.890 30	0.636 31	0.516 64	0.693 55	0.743 26	0.588 49	0.369 60	0.601 47	0.594 68	0.800 51	0.886 16	0.676 56	0.986 53	0.546 57

SegGroup_ins	0.637 63	1.000 1	0.923 48	0.593 68	0.561 66	0.746 61	0.143 73	0.504 65	0.766 45	0.485 67	0.442 65	0.372 59	0.530 58	0.714 63	0.815 49	0.775 65	0.673 57	1.000 1	0.431 66
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
3D-BoNet	0.687 58	1.000 1	0.887 58	0.836 29	0.587 65	0.643 72	0.550 43	0.620 53	0.724 50	0.522 64	0.501 62	0.243 64	0.512 60	1.000 1	0.751 58	0.807 59	0.661 58	0.909 67	0.612 51
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
PanopticFusion-inst	0.693 56	1.000 1	0.852 62	0.655 63	0.616 61	0.788 55	0.334 59	0.763 34	0.771 43	0.457 69	0.555 55	0.652 30	0.518 59	0.857 54	0.765 56	0.732 69	0.631 59	0.944 58	0.577 55
Gaku Narita, Takashi Seno, Tomoya Ishikawa, Yohsuke Kaji: PanopticFusion: Online Volumetric Semantic Mapping at the Level of Stuff and Things. IROS 2019 (to appear)
PCJC	0.684 60	1.000 1	0.895 57	0.757 53	0.659 58	0.862 39	0.189 71	0.739 40	0.606 61	0.712 32	0.581 50	0.515 50	0.650 31	0.857 54	0.357 72	0.785 64	0.631 60	0.889 68	0.635 45

OSIS	0.725 51	1.000 1	0.885 59	0.653 64	0.657 59	0.801 53	0.576 37	0.695 46	0.828 31	0.698 33	0.534 58	0.457 55	0.500 62	0.857 54	0.831 45	0.841 47	0.627 61	1.000 1	0.619 47

ClickSeg_Instance	0.685 59	1.000 1	0.818 65	0.600 67	0.715 52	0.795 54	0.557 39	0.533 62	0.591 64	0.601 49	0.519 60	0.429 58	0.638 44	0.938 49	0.706 61	0.817 55	0.624 62	0.944 58	0.502 62

OccuSeg+instance	0.742 47	1.000 1	0.923 46	0.785 43	0.745 44	0.867 34	0.557 39	0.578 60	0.729 49	0.670 39	0.644 39	0.488 51	0.577 54	1.000 1	0.794 53	0.830 50	0.620 63	1.000 1	0.550 56
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
R-PointNet	0.544 67	0.500 74	0.655 73	0.661 62	0.663 56	0.765 58	0.432 54	0.214 73	0.612 60	0.584 54	0.499 63	0.204 66	0.286 72	0.429 71	0.655 63	0.650 74	0.539 64	0.950 57	0.499 63

MASC	0.615 64	0.711 71	0.802 66	0.540 70	0.757 38	0.777 56	0.029 74	0.577 61	0.588 65	0.521 65	0.600 46	0.436 57	0.534 57	0.697 64	0.616 65	0.838 48	0.526 65	0.980 55	0.534 59
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
Sem_Recon_ins	0.484 70	0.764 70	0.608 75	0.470 72	0.521 68	0.637 73	0.311 60	0.218 72	0.348 74	0.365 73	0.223 72	0.222 65	0.258 73	0.629 67	0.734 59	0.596 75	0.509 66	0.858 71	0.444 65

UNet-backbone	0.605 65	1.000 1	0.909 51	0.764 51	0.603 63	0.704 66	0.415 55	0.301 70	0.548 67	0.461 68	0.394 66	0.267 62	0.386 68	0.857 54	0.649 64	0.817 54	0.504 67	0.959 56	0.356 70

tmp	0.474 71	1.000 1	0.727 68	0.433 74	0.481 71	0.673 70	0.022 76	0.380 68	0.517 69	0.436 71	0.338 69	0.128 69	0.343 70	0.429 71	0.291 74	0.728 70	0.473 68	0.833 72	0.300 73

Hier3D	0.540 68	1.000 1	0.727 68	0.626 65	0.467 72	0.693 67	0.200 69	0.412 66	0.480 71	0.528 63	0.318 70	0.077 74	0.600 48	0.688 65	0.382 69	0.768 67	0.472 69	0.941 62	0.350 71
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
SemRegionNet-20cls	0.470 72	1.000 1	0.727 68	0.447 73	0.481 70	0.678 69	0.024 75	0.380 68	0.518 68	0.440 70	0.339 68	0.128 69	0.350 69	0.429 71	0.212 75	0.711 71	0.465 70	0.833 72	0.290 74

Region-18class	0.497 69	0.250 76	0.902 52	0.689 60	0.540 67	0.747 59	0.276 65	0.610 54	0.268 75	0.489 66	0.348 67	0.000 75	0.243 75	0.220 74	0.663 62	0.814 56	0.459 71	0.928 66	0.496 64

ASIS	0.422 73	0.333 75	0.707 71	0.676 61	0.401 73	0.650 71	0.350 58	0.177 74	0.594 63	0.376 72	0.202 73	0.077 73	0.404 67	0.571 69	0.197 76	0.674 73	0.447 72	0.500 75	0.260 75

3D-SIS	0.558 66	1.000 1	0.773 67	0.614 66	0.503 69	0.691 68	0.200 69	0.412 66	0.498 70	0.546 61	0.311 71	0.103 71	0.600 48	0.857 54	0.382 69	0.799 62	0.445 73	0.938 64	0.371 68
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
Sgpn_scannet	0.390 75	0.556 73	0.636 74	0.493 71	0.353 74	0.539 75	0.271 67	0.160 75	0.450 72	0.359 74	0.178 74	0.146 68	0.250 74	0.143 75	0.347 73	0.698 72	0.436 74	0.667 74	0.331 72

3D-BEVIS	0.401 74	0.667 72	0.687 72	0.419 75	0.137 76	0.587 74	0.188 72	0.235 71	0.359 73	0.211 75	0.093 76	0.080 72	0.311 71	0.571 69	0.382 69	0.754 68	0.300 75	0.874 70	0.357 69
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
MaskRCNN 2d->3d Proj	0.261 76	0.903 69	0.081 76	0.008 76	0.233 75	0.175 76	0.280 63	0.106 76	0.150 76	0.203 76	0.175 75	0.480 52	0.218 76	0.143 75	0.542 68	0.404 76	0.153 76	0.393 76	0.049 76

3D Semantic Instance Benchmark