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 50%	bathtub	bed	bookshelf	cabinet	chair	counter	curtain	desk	door	otherfurniture	picture	refrigerator	shower curtain	sink	sofa	table	toilet	window

AQ3D	0.834 1	1.000 1	0.932 15	0.876 12	0.752 3	0.948 1	0.632 2	0.680 32	0.761 12	0.784 2	0.753 3	0.826 1	0.812 2	1.000 1	0.859 2	0.879 8	0.831 2	1.000 1	0.688 3

Volt-SPFormerScanNet	0.827 2	1.000 1	0.981 6	0.975 1	0.801 1	0.940 5	0.426 25	0.693 30	0.752 14	0.762 8	0.800 1	0.804 3	0.855 1	0.959 49	0.745 24	0.879 7	0.806 8	0.997 44	0.710 1
Kadir Yilmaz, Adrian Kruse, Tristan Höfer, Daan de Geus, Bastian Leibe: Volume Transformer: Revisiting Vanilla Transformers for 3D Scene Understanding.
Competitor-MAFT	0.816 3	1.000 1	0.983 4	0.872 13	0.718 7	0.941 4	0.588 6	0.652 43	0.819 3	0.776 4	0.720 8	0.780 8	0.769 13	1.000 1	0.797 12	0.813 33	0.798 10	1.000 1	0.659 6

PointRel	0.816 3	1.000 1	0.971 10	0.908 7	0.743 4	0.923 12	0.573 10	0.714 22	0.695 22	0.734 12	0.747 4	0.725 15	0.809 3	1.000 1	0.814 10	0.899 5	0.820 4	1.000 1	0.610 21
: Relation3D: Enhancing Relation Modeling for Point Cloud Instance Segmentation. CVPR 2025
Spherical Mask(CtoF)	0.812 5	1.000 1	0.973 9	0.852 17	0.718 8	0.917 14	0.574 8	0.677 33	0.748 15	0.729 16	0.715 11	0.795 5	0.809 3	1.000 1	0.831 5	0.854 14	0.787 14	1.000 1	0.638 10

EV3D	0.811 6	1.000 1	0.968 12	0.852 17	0.717 9	0.921 13	0.574 9	0.677 33	0.748 15	0.730 15	0.703 17	0.795 5	0.809 3	1.000 1	0.831 5	0.854 14	0.778 18	1.000 1	0.638 11

PointComp	0.811 6	0.850 62	0.969 11	0.864 15	0.739 5	0.946 3	0.539 17	0.671 36	0.835 2	0.700 20	0.742 5	0.817 2	0.766 14	1.000 1	0.755 22	0.909 1	0.808 7	1.000 1	0.687 4

VDG-Uni3DSeg	0.804 8	1.000 1	0.990 1	0.886 10	0.688 22	0.912 16	0.602 3	0.703 26	0.786 8	0.771 5	0.708 15	0.700 20	0.669 28	0.981 42	0.789 18	0.903 2	0.772 22	1.000 1	0.609 22

SIM3D	0.803 9	1.000 1	0.967 13	0.863 16	0.692 21	0.924 11	0.552 14	0.732 20	0.667 27	0.732 14	0.662 21	0.796 4	0.789 11	1.000 1	0.803 11	0.864 11	0.766 25	1.000 1	0.643 8

OneFormer3D	0.801 10	1.000 1	0.973 8	0.909 6	0.698 17	0.928 9	0.582 7	0.668 39	0.685 23	0.780 3	0.687 19	0.698 24	0.702 17	1.000 1	0.794 14	0.900 4	0.784 16	0.986 57	0.635 12
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
Competitor-SPFormer	0.800 11	1.000 1	0.986 3	0.845 19	0.705 15	0.915 15	0.532 18	0.733 19	0.757 13	0.733 13	0.708 14	0.698 23	0.648 40	0.981 42	0.890 1	0.830 24	0.796 11	0.997 44	0.644 7

InsSSM	0.799 12	1.000 1	0.915 17	0.710 46	0.729 6	0.925 10	0.664 1	0.670 37	0.770 9	0.766 6	0.739 6	0.737 11	0.700 18	1.000 1	0.792 15	0.829 26	0.815 5	0.997 44	0.625 14
Lei Yao, Yi Wang, Moyun Liu, Lap-Pui Chau: SGIFormer: Semantic-guided and Geometric-enhanced Interleaving Transformer for 3D Instance Segmentation. TCSVT, 2024
DCD	0.798 13	1.000 1	0.878 25	0.792 31	0.693 20	0.936 6	0.596 4	0.685 31	0.663 29	0.736 10	0.717 9	0.788 7	0.693 23	1.000 1	0.825 8	0.840 20	0.837 1	1.000 1	0.689 2

TST3D	0.795 14	1.000 1	0.929 16	0.918 5	0.709 12	0.884 25	0.596 5	0.704 25	0.769 10	0.734 11	0.644 26	0.699 22	0.751 15	1.000 1	0.794 13	0.876 10	0.757 28	0.997 44	0.550 38
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
MG-Former	0.791 15	1.000 1	0.980 7	0.837 22	0.626 31	0.897 18	0.543 16	0.759 14	0.800 7	0.766 7	0.659 22	0.769 10	0.697 21	1.000 1	0.791 16	0.707 54	0.791 13	1.000 1	0.610 20

ExtMask3D	0.789 16	1.000 1	0.988 2	0.756 39	0.706 14	0.912 17	0.429 24	0.647 45	0.806 6	0.755 9	0.673 20	0.689 25	0.772 12	1.000 1	0.789 17	0.852 16	0.811 6	1.000 1	0.617 17

UniPerception	0.787 17	1.000 1	0.909 18	0.768 36	0.687 23	0.947 2	0.551 15	0.714 21	0.843 1	0.696 21	0.713 13	0.773 9	0.607 46	0.981 42	0.690 31	0.878 9	0.775 21	1.000 1	0.640 9

Queryformer	0.787 17	1.000 1	0.933 14	0.601 56	0.754 2	0.886 23	0.558 13	0.661 41	0.767 11	0.665 24	0.716 10	0.639 31	0.808 7	1.000 1	0.844 4	0.897 6	0.804 9	1.000 1	0.624 15

MAFT	0.786 19	1.000 1	0.894 23	0.807 26	0.694 19	0.893 21	0.486 20	0.674 35	0.740 17	0.786 1	0.704 16	0.727 14	0.739 16	1.000 1	0.707 29	0.849 18	0.756 29	1.000 1	0.685 5

KmaxOneFormerNet	0.783 20	0.903 60	0.981 5	0.794 30	0.706 13	0.931 8	0.561 12	0.701 27	0.706 20	0.727 17	0.697 18	0.731 13	0.689 25	1.000 1	0.856 3	0.750 45	0.761 27	1.000 1	0.599 26

Mask3D	0.780 21	1.000 1	0.786 49	0.716 44	0.696 18	0.885 24	0.500 19	0.714 22	0.810 5	0.672 23	0.715 11	0.679 26	0.809 3	1.000 1	0.831 5	0.833 23	0.787 14	1.000 1	0.602 24
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
SPFormer	0.770 22	0.903 60	0.903 20	0.806 27	0.609 38	0.886 22	0.568 11	0.815 6	0.705 21	0.711 18	0.655 23	0.652 30	0.685 26	1.000 1	0.789 19	0.809 34	0.776 20	1.000 1	0.583 30
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
SoftGroup++	0.769 23	1.000 1	0.803 42	0.937 2	0.684 24	0.865 27	0.213 41	0.870 2	0.664 28	0.571 31	0.758 2	0.702 19	0.807 8	1.000 1	0.653 37	0.902 3	0.792 12	1.000 1	0.626 13

SoftGroup	0.761 24	1.000 1	0.808 38	0.845 19	0.716 10	0.862 29	0.243 38	0.824 4	0.655 31	0.620 25	0.734 7	0.699 21	0.791 10	0.981 42	0.716 26	0.844 19	0.769 23	1.000 1	0.594 28
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral]
ISBNet	0.757 25	1.000 1	0.904 19	0.731 42	0.678 25	0.895 19	0.458 22	0.644 47	0.670 26	0.710 19	0.620 31	0.732 12	0.650 30	1.000 1	0.756 21	0.778 37	0.779 17	1.000 1	0.614 18
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
TD3D	0.751 26	1.000 1	0.774 50	0.867 14	0.621 33	0.934 7	0.404 26	0.706 24	0.812 4	0.605 28	0.633 29	0.626 32	0.690 24	1.000 1	0.640 39	0.820 29	0.777 19	1.000 1	0.612 19
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
PBNet	0.747 27	1.000 1	0.818 34	0.837 23	0.713 11	0.844 31	0.457 23	0.647 45	0.711 19	0.614 26	0.617 33	0.657 29	0.650 30	1.000 1	0.692 30	0.822 28	0.765 26	1.000 1	0.595 27
Weiguang Zhao, Yuyao Yan, Chaolong Yang, Jianan Ye, Xi Yang, Kaizhu Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
GraphCut	0.732 28	1.000 1	0.788 47	0.724 43	0.642 30	0.859 30	0.248 37	0.787 11	0.618 34	0.596 29	0.653 25	0.722 17	0.583 53	1.000 1	0.766 20	0.861 12	0.825 3	1.000 1	0.504 44

IPCA-Inst	0.731 29	1.000 1	0.788 48	0.884 11	0.698 16	0.788 47	0.252 36	0.760 13	0.646 32	0.511 39	0.637 28	0.665 28	0.804 9	1.000 1	0.644 38	0.778 38	0.747 31	1.000 1	0.561 34

TopoSeg	0.725 30	1.000 1	0.806 41	0.933 3	0.668 27	0.758 52	0.272 35	0.734 18	0.630 33	0.549 35	0.654 24	0.606 33	0.697 22	0.966 48	0.612 43	0.839 21	0.754 30	1.000 1	0.573 31

DKNet	0.718 31	1.000 1	0.814 35	0.782 32	0.619 35	0.872 26	0.224 39	0.751 16	0.569 38	0.677 22	0.585 38	0.724 16	0.633 42	0.981 42	0.515 53	0.819 30	0.736 32	1.000 1	0.617 16
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
SSEC	0.707 32	1.000 1	0.850 27	0.924 4	0.648 28	0.747 55	0.162 43	0.862 3	0.572 37	0.520 37	0.624 30	0.549 36	0.649 39	1.000 1	0.560 48	0.706 55	0.768 24	1.000 1	0.591 29

HAIS	0.699 33	1.000 1	0.849 28	0.820 24	0.675 26	0.808 41	0.279 33	0.757 15	0.465 44	0.517 38	0.596 35	0.559 35	0.600 47	1.000 1	0.654 36	0.767 40	0.676 36	0.994 53	0.560 35
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
SSTNet	0.698 34	1.000 1	0.697 66	0.888 9	0.556 45	0.803 42	0.387 27	0.626 49	0.417 49	0.556 34	0.585 39	0.702 18	0.600 47	1.000 1	0.824 9	0.720 53	0.692 34	1.000 1	0.509 43
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
DualGroup	0.694 35	1.000 1	0.799 44	0.811 25	0.622 32	0.817 36	0.376 28	0.805 9	0.590 36	0.487 43	0.568 42	0.525 40	0.650 30	0.835 61	0.600 44	0.829 25	0.655 39	1.000 1	0.526 40

ODIN - Ins	0.693 36	1.000 1	0.880 24	0.647 51	0.620 34	0.779 49	0.336 30	0.501 64	0.681 24	0.577 30	0.595 36	0.679 27	0.683 27	1.000 1	0.709 28	0.816 32	0.637 43	0.770 73	0.557 36
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
SphereSeg	0.680 37	1.000 1	0.856 26	0.744 40	0.618 36	0.893 20	0.151 44	0.651 44	0.713 18	0.537 36	0.579 41	0.430 50	0.651 29	1.000 1	0.389 64	0.744 48	0.697 33	0.991 55	0.601 25

DANCENET	0.680 37	1.000 1	0.807 39	0.733 41	0.600 39	0.768 51	0.375 29	0.543 57	0.538 39	0.610 27	0.599 34	0.498 41	0.632 44	0.981 42	0.739 25	0.856 13	0.633 46	0.882 68	0.454 53

Box2Mask	0.677 39	1.000 1	0.847 29	0.771 34	0.509 54	0.816 37	0.277 34	0.558 56	0.482 41	0.562 33	0.640 27	0.448 46	0.700 18	1.000 1	0.666 32	0.852 17	0.578 53	0.997 44	0.488 48
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022
OccuSeg+instance	0.672 40	1.000 1	0.758 58	0.682 48	0.576 43	0.842 32	0.477 21	0.504 63	0.524 40	0.567 32	0.585 40	0.451 45	0.557 55	1.000 1	0.751 23	0.797 35	0.563 56	1.000 1	0.467 52
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
Mask-Group	0.664 41	1.000 1	0.822 33	0.764 38	0.616 37	0.815 38	0.139 48	0.694 29	0.597 35	0.459 47	0.566 43	0.599 34	0.600 47	0.516 71	0.715 27	0.819 31	0.635 44	1.000 1	0.603 23
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022
INS-Conv-instance	0.657 42	1.000 1	0.760 56	0.667 50	0.581 41	0.863 28	0.323 31	0.655 42	0.477 42	0.473 45	0.549 45	0.432 49	0.650 30	1.000 1	0.655 35	0.738 49	0.585 52	0.944 60	0.472 51

CSC-Pretrained	0.648 43	1.000 1	0.810 36	0.768 35	0.523 52	0.813 39	0.143 47	0.819 5	0.389 52	0.422 56	0.511 49	0.443 47	0.650 30	1.000 1	0.624 41	0.732 50	0.634 45	1.000 1	0.375 60

PE	0.645 44	1.000 1	0.773 52	0.798 29	0.538 47	0.786 48	0.088 56	0.799 10	0.350 56	0.435 54	0.547 46	0.545 37	0.646 41	0.933 51	0.562 47	0.761 43	0.556 61	0.997 44	0.501 46
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
RPGN	0.643 45	1.000 1	0.758 57	0.582 62	0.539 46	0.826 35	0.046 61	0.765 12	0.372 54	0.436 53	0.588 37	0.539 39	0.650 30	1.000 1	0.577 45	0.750 46	0.653 41	0.997 44	0.495 47
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
Dyco3D	0.641 46	1.000 1	0.841 30	0.893 8	0.531 49	0.802 43	0.115 53	0.588 54	0.448 46	0.438 51	0.537 48	0.430 51	0.550 56	0.857 53	0.534 51	0.764 42	0.657 38	0.987 56	0.568 32
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
GICN	0.638 47	1.000 1	0.895 22	0.800 28	0.480 58	0.676 60	0.144 46	0.737 17	0.354 55	0.447 48	0.400 62	0.365 57	0.700 18	1.000 1	0.569 46	0.836 22	0.599 48	1.000 1	0.473 50

PointGroup	0.636 48	1.000 1	0.765 53	0.624 53	0.505 56	0.797 44	0.116 52	0.696 28	0.384 53	0.441 49	0.559 44	0.476 43	0.596 50	1.000 1	0.666 32	0.756 44	0.556 60	0.997 44	0.513 42
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]
DD-UNet+Group	0.635 49	0.667 64	0.797 46	0.714 45	0.562 44	0.774 50	0.146 45	0.810 8	0.429 48	0.476 44	0.546 47	0.399 53	0.633 42	1.000 1	0.632 40	0.722 52	0.609 47	1.000 1	0.514 41
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
Mask3D_evaluation	0.631 50	1.000 1	0.829 32	0.606 55	0.646 29	0.836 33	0.068 57	0.511 61	0.462 45	0.507 40	0.619 32	0.389 55	0.610 45	1.000 1	0.432 59	0.828 27	0.673 37	0.788 72	0.552 37

DENet	0.629 51	1.000 1	0.797 45	0.608 54	0.589 40	0.627 64	0.219 40	0.882 1	0.310 58	0.402 61	0.383 64	0.396 54	0.650 30	1.000 1	0.663 34	0.543 72	0.691 35	1.000 1	0.568 33

3D-MPA	0.611 52	1.000 1	0.833 31	0.765 37	0.526 51	0.756 53	0.136 50	0.588 54	0.470 43	0.438 52	0.432 58	0.358 59	0.650 30	0.857 53	0.429 60	0.765 41	0.557 59	1.000 1	0.430 55
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020
OSIS	0.605 53	1.000 1	0.801 43	0.599 57	0.535 48	0.728 57	0.286 32	0.436 68	0.679 25	0.491 41	0.433 56	0.256 61	0.404 68	0.857 53	0.620 42	0.724 51	0.510 66	1.000 1	0.539 39

AOIA	0.601 54	1.000 1	0.761 55	0.687 47	0.485 57	0.828 34	0.008 68	0.663 40	0.405 51	0.405 60	0.425 59	0.490 42	0.596 50	0.714 64	0.553 50	0.779 36	0.597 49	0.992 54	0.424 57

PCJC	0.578 55	1.000 1	0.810 37	0.583 61	0.449 61	0.813 40	0.042 62	0.603 52	0.341 57	0.490 42	0.465 53	0.410 52	0.650 30	0.835 61	0.264 70	0.694 59	0.561 57	0.889 65	0.504 45

SSEN	0.575 56	1.000 1	0.761 54	0.473 64	0.477 59	0.795 45	0.066 58	0.529 59	0.658 30	0.460 46	0.461 54	0.380 56	0.331 70	0.859 52	0.401 63	0.692 61	0.653 40	1.000 1	0.348 62
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv
RWSeg	0.567 57	0.528 74	0.708 65	0.626 52	0.580 42	0.745 56	0.063 59	0.627 48	0.240 62	0.400 62	0.497 50	0.464 44	0.515 57	1.000 1	0.475 55	0.745 47	0.571 54	1.000 1	0.429 56

NeuralBF	0.555 58	0.667 64	0.896 21	0.843 21	0.517 53	0.751 54	0.029 63	0.519 60	0.414 50	0.439 50	0.465 52	0.000 80	0.484 59	0.857 53	0.287 68	0.693 60	0.651 42	1.000 1	0.485 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
MTML	0.549 59	1.000 1	0.807 40	0.588 60	0.327 66	0.647 62	0.004 70	0.815 7	0.180 65	0.418 57	0.364 66	0.182 64	0.445 62	1.000 1	0.442 58	0.688 62	0.571 55	1.000 1	0.396 58
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
ClickSeg_Instance	0.539 60	1.000 1	0.621 69	0.300 67	0.530 50	0.698 58	0.127 51	0.533 58	0.222 63	0.430 55	0.400 61	0.365 57	0.574 54	0.938 50	0.472 56	0.659 64	0.543 62	0.944 60	0.347 63

One_Thing_One_Click	0.529 61	0.667 64	0.718 61	0.777 33	0.399 62	0.683 59	0.000 73	0.669 38	0.138 68	0.391 63	0.374 65	0.539 38	0.360 69	0.641 68	0.556 49	0.774 39	0.593 50	0.997 44	0.251 68
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
Sparse R-CNN	0.515 62	1.000 1	0.538 74	0.282 68	0.468 60	0.790 46	0.173 42	0.345 70	0.429 47	0.413 59	0.484 51	0.176 65	0.595 52	0.591 69	0.522 52	0.668 63	0.476 67	0.986 58	0.327 64

Occipital-SCS	0.512 63	1.000 1	0.716 62	0.509 63	0.506 55	0.611 65	0.092 55	0.602 53	0.177 66	0.346 66	0.383 63	0.165 66	0.442 63	0.850 60	0.386 65	0.618 68	0.543 63	0.889 65	0.389 59

3D-BoNet	0.488 64	1.000 1	0.672 68	0.590 59	0.301 68	0.484 75	0.098 54	0.620 50	0.306 59	0.341 67	0.259 70	0.125 68	0.434 65	0.796 63	0.402 62	0.499 74	0.513 65	0.909 64	0.439 54
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	0.478 65	0.667 64	0.712 64	0.595 58	0.259 71	0.550 71	0.000 73	0.613 51	0.175 67	0.250 72	0.434 55	0.437 48	0.411 67	0.857 53	0.485 54	0.591 71	0.267 77	0.944 60	0.359 61
Gaku Narita, Takashi Seno, Tomoya Ishikawa, Yohsuke Kaji: PanopticFusion: Online Volumetric Semantic Mapping at the Level of Stuff and Things. IROS 2019 (to appear)
SPG_WSIS	0.470 66	0.667 64	0.685 67	0.677 49	0.372 64	0.562 69	0.000 73	0.482 65	0.244 61	0.316 69	0.298 67	0.052 75	0.442 64	0.857 53	0.267 69	0.702 56	0.559 58	1.000 1	0.287 66

SALoss-ResNet	0.459 67	1.000 1	0.737 60	0.159 78	0.259 70	0.587 67	0.138 49	0.475 66	0.217 64	0.416 58	0.408 60	0.128 67	0.315 71	0.714 64	0.411 61	0.536 73	0.590 51	0.873 69	0.304 65
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)
MASC	0.447 68	0.528 74	0.555 72	0.381 65	0.382 63	0.633 63	0.002 71	0.509 62	0.260 60	0.361 65	0.432 57	0.327 60	0.451 61	0.571 70	0.367 66	0.639 66	0.386 68	0.980 59	0.276 67
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
SegGroup_ins	0.445 69	0.667 64	0.773 51	0.185 75	0.317 67	0.656 61	0.000 73	0.407 69	0.134 69	0.381 64	0.267 69	0.217 63	0.476 60	0.714 64	0.452 57	0.629 67	0.514 64	1.000 1	0.222 71
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
3D-SIS	0.382 70	1.000 1	0.432 77	0.245 70	0.190 72	0.577 68	0.013 67	0.263 72	0.033 75	0.320 68	0.240 71	0.075 71	0.422 66	0.857 53	0.117 75	0.699 57	0.271 76	0.883 67	0.235 70
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
Hier3D	0.323 71	0.667 64	0.542 73	0.264 69	0.157 75	0.550 70	0.000 73	0.205 75	0.009 77	0.270 71	0.218 72	0.075 71	0.500 58	0.688 67	0.007 81	0.698 58	0.301 73	0.459 78	0.200 72
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
UNet-backbone	0.319 72	0.667 64	0.715 63	0.233 71	0.189 73	0.479 76	0.008 68	0.218 73	0.067 74	0.201 74	0.173 73	0.107 69	0.123 76	0.438 72	0.150 72	0.615 69	0.355 69	0.916 63	0.093 80

R-PointNet	0.306 73	0.500 76	0.405 78	0.311 66	0.348 65	0.589 66	0.054 60	0.068 78	0.126 70	0.283 70	0.290 68	0.028 76	0.219 74	0.214 75	0.331 67	0.396 78	0.275 74	0.821 71	0.245 69

Region-18class	0.284 74	0.250 80	0.751 59	0.228 73	0.270 69	0.521 72	0.000 73	0.468 67	0.008 79	0.205 73	0.127 74	0.000 80	0.068 78	0.070 79	0.262 71	0.652 65	0.323 71	0.740 74	0.173 73

SemRegionNet-20cls	0.250 75	0.333 77	0.613 70	0.229 72	0.163 74	0.493 73	0.000 73	0.304 71	0.107 71	0.147 77	0.100 76	0.052 74	0.231 72	0.119 77	0.039 77	0.445 76	0.325 70	0.654 75	0.141 76

tmp	0.248 76	0.667 64	0.437 76	0.188 74	0.153 76	0.491 74	0.000 73	0.208 74	0.094 73	0.153 76	0.099 77	0.057 73	0.217 75	0.119 77	0.039 77	0.466 75	0.302 72	0.640 76	0.140 77

3D-BEVIS	0.248 76	0.667 64	0.566 71	0.076 79	0.035 81	0.394 79	0.027 65	0.035 80	0.098 72	0.099 79	0.030 80	0.025 77	0.098 77	0.375 74	0.126 74	0.604 70	0.181 79	0.854 70	0.171 74
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
Sem_Recon_ins	0.227 78	0.764 63	0.486 75	0.069 80	0.098 78	0.426 78	0.017 66	0.067 79	0.015 76	0.172 75	0.100 75	0.096 70	0.054 80	0.183 76	0.135 73	0.366 79	0.260 78	0.614 77	0.168 75

ASIS	0.199 79	0.333 77	0.253 80	0.167 77	0.140 77	0.438 77	0.000 73	0.177 76	0.008 78	0.121 78	0.069 78	0.004 79	0.231 73	0.429 73	0.036 79	0.445 77	0.273 75	0.333 80	0.119 79

Sgpn_scannet	0.143 80	0.208 81	0.390 79	0.169 76	0.065 79	0.275 80	0.029 64	0.069 77	0.000 80	0.087 80	0.043 79	0.014 78	0.027 81	0.000 80	0.112 76	0.351 80	0.168 80	0.438 79	0.138 78

MaskRCNN 2d->3d Proj	0.058 81	0.333 77	0.002 81	0.000 81	0.053 80	0.002 81	0.002 72	0.021 81	0.000 80	0.045 81	0.024 81	0.238 62	0.065 79	0.000 80	0.014 80	0.107 81	0.020 81	0.110 81	0.006 81

3D Semantic Instance Benchmark