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

Competitor-MAFT	0.816 1	1.000 1	0.983 4	0.872 11	0.718 5	0.941 1	0.588 5	0.652 40	0.819 2	0.776 3	0.720 5	0.780 5	0.769 12	1.000 1	0.797 11	0.813 30	0.798 8	1.000 1	0.659 4

PointRel	0.816 1	1.000 1	0.971 9	0.908 6	0.743 2	0.923 8	0.573 9	0.714 22	0.695 19	0.734 11	0.747 2	0.725 12	0.809 1	1.000 1	0.814 9	0.899 4	0.820 4	1.000 1	0.610 18
: Relation3D: Enhancing Relation Modeling for Point Cloud Instance Segmentation. CVPR 2025
Spherical Mask(CtoF)	0.812 3	1.000 1	0.973 8	0.852 15	0.718 6	0.917 10	0.574 7	0.677 31	0.748 12	0.729 15	0.715 8	0.795 2	0.809 1	1.000 1	0.831 4	0.854 10	0.787 12	1.000 1	0.638 7

EV3D	0.811 4	1.000 1	0.968 10	0.852 15	0.717 7	0.921 9	0.574 8	0.677 31	0.748 12	0.730 14	0.703 14	0.795 2	0.809 1	1.000 1	0.831 4	0.854 10	0.778 16	1.000 1	0.638 8

VDG-Uni3DSeg	0.804 5	1.000 1	0.990 1	0.886 9	0.688 20	0.912 12	0.602 2	0.703 26	0.786 7	0.771 4	0.708 11	0.700 17	0.669 26	0.981 40	0.789 17	0.903 1	0.772 19	1.000 1	0.609 19

SIM3D	0.803 6	1.000 1	0.967 11	0.863 14	0.692 19	0.924 7	0.552 13	0.732 21	0.667 24	0.732 13	0.662 18	0.796 1	0.789 9	1.000 1	0.803 10	0.864 7	0.766 22	1.000 1	0.643 6

OneFormer3D	0.801 7	1.000 1	0.973 7	0.909 5	0.698 15	0.928 5	0.582 6	0.668 36	0.685 20	0.780 2	0.687 16	0.698 21	0.702 15	1.000 1	0.794 13	0.900 3	0.784 14	0.986 54	0.635 9
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
Competitor-SPFormer	0.800 8	1.000 1	0.986 3	0.845 17	0.705 13	0.915 11	0.532 15	0.733 20	0.757 11	0.733 12	0.708 10	0.698 20	0.648 38	0.981 40	0.890 1	0.830 20	0.796 9	0.997 41	0.644 5

UniPerception	0.800 8	1.000 1	0.930 13	0.872 11	0.727 4	0.862 26	0.454 21	0.764 13	0.820 1	0.746 8	0.706 12	0.750 7	0.772 10	0.926 48	0.764 20	0.818 28	0.826 2	0.997 41	0.660 3

InsSSM	0.799 10	1.000 1	0.915 15	0.710 43	0.729 3	0.925 6	0.664 1	0.670 34	0.770 8	0.766 5	0.739 3	0.737 8	0.700 16	1.000 1	0.792 14	0.829 22	0.815 5	0.997 41	0.625 11
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 11	1.000 1	0.878 22	0.792 29	0.693 18	0.936 2	0.596 3	0.685 30	0.663 26	0.736 9	0.717 6	0.788 4	0.693 21	1.000 1	0.825 7	0.840 16	0.837 1	1.000 1	0.689 1

TST3D	0.795 12	1.000 1	0.929 14	0.918 4	0.709 10	0.884 21	0.596 4	0.704 25	0.769 9	0.734 10	0.644 23	0.699 19	0.751 13	1.000 1	0.794 12	0.876 6	0.757 25	0.997 41	0.550 35
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
MG-Former	0.791 13	1.000 1	0.980 6	0.837 20	0.626 28	0.897 14	0.543 14	0.759 15	0.800 6	0.766 6	0.659 19	0.769 6	0.697 19	1.000 1	0.791 15	0.707 51	0.791 11	1.000 1	0.610 17

ExtMask3D	0.789 14	1.000 1	0.988 2	0.756 36	0.706 12	0.912 13	0.429 22	0.647 42	0.806 5	0.755 7	0.673 17	0.689 22	0.772 11	1.000 1	0.789 16	0.852 12	0.811 6	1.000 1	0.617 14

Queryformer	0.787 15	1.000 1	0.933 12	0.601 53	0.754 1	0.886 19	0.558 12	0.661 38	0.767 10	0.665 21	0.716 7	0.639 28	0.808 5	1.000 1	0.844 3	0.897 5	0.804 7	1.000 1	0.624 12

MAFT	0.786 16	1.000 1	0.894 20	0.807 24	0.694 17	0.893 17	0.486 17	0.674 33	0.740 14	0.786 1	0.704 13	0.727 11	0.739 14	1.000 1	0.707 27	0.849 14	0.756 26	1.000 1	0.685 2

KmaxOneFormerNet	0.783 17	0.903 58	0.981 5	0.794 28	0.706 11	0.931 4	0.561 11	0.701 27	0.706 17	0.727 16	0.697 15	0.731 10	0.689 23	1.000 1	0.856 2	0.750 42	0.761 24	1.000 1	0.599 23

Mask3D	0.780 18	1.000 1	0.786 46	0.716 41	0.696 16	0.885 20	0.500 16	0.714 22	0.810 4	0.672 20	0.715 8	0.679 23	0.809 1	1.000 1	0.831 4	0.833 19	0.787 12	1.000 1	0.602 21
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
SPFormer	0.770 19	0.903 58	0.903 17	0.806 25	0.609 35	0.886 18	0.568 10	0.815 6	0.705 18	0.711 17	0.655 20	0.652 27	0.685 24	1.000 1	0.789 18	0.809 31	0.776 18	1.000 1	0.583 27
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
SoftGroup++	0.769 20	1.000 1	0.803 39	0.937 1	0.684 21	0.865 23	0.213 38	0.870 2	0.664 25	0.571 28	0.758 1	0.702 16	0.807 6	1.000 1	0.653 34	0.902 2	0.792 10	1.000 1	0.626 10

SoftGroup	0.761 21	1.000 1	0.808 35	0.845 17	0.716 8	0.862 25	0.243 35	0.824 4	0.655 28	0.620 22	0.734 4	0.699 18	0.791 8	0.981 40	0.716 24	0.844 15	0.769 20	1.000 1	0.594 25
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 22	1.000 1	0.904 16	0.731 39	0.678 22	0.895 15	0.458 19	0.644 44	0.670 23	0.710 18	0.620 28	0.732 9	0.650 28	1.000 1	0.756 21	0.778 34	0.779 15	1.000 1	0.614 15
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 23	1.000 1	0.774 47	0.867 13	0.621 30	0.934 3	0.404 23	0.706 24	0.812 3	0.605 25	0.633 26	0.626 29	0.690 22	1.000 1	0.640 36	0.820 25	0.777 17	1.000 1	0.612 16
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
PBNet	0.747 24	1.000 1	0.818 31	0.837 21	0.713 9	0.844 28	0.457 20	0.647 42	0.711 16	0.614 23	0.617 30	0.657 26	0.650 28	1.000 1	0.692 28	0.822 24	0.765 23	1.000 1	0.595 24
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 25	1.000 1	0.788 44	0.724 40	0.642 27	0.859 27	0.248 34	0.787 11	0.618 31	0.596 26	0.653 22	0.722 14	0.583 50	1.000 1	0.766 19	0.861 8	0.825 3	1.000 1	0.504 41

IPCA-Inst	0.731 26	1.000 1	0.788 45	0.884 10	0.698 14	0.788 44	0.252 33	0.760 14	0.646 29	0.511 36	0.637 25	0.665 25	0.804 7	1.000 1	0.644 35	0.778 35	0.747 28	1.000 1	0.561 31

TopoSeg	0.725 27	1.000 1	0.806 38	0.933 2	0.668 24	0.758 49	0.272 32	0.734 19	0.630 30	0.549 32	0.654 21	0.606 30	0.697 20	0.966 45	0.612 40	0.839 17	0.754 27	1.000 1	0.573 28

DKNet	0.718 28	1.000 1	0.814 32	0.782 30	0.619 32	0.872 22	0.224 36	0.751 17	0.569 35	0.677 19	0.585 35	0.724 13	0.633 40	0.981 40	0.515 50	0.819 26	0.736 29	1.000 1	0.617 13
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
SSEC	0.707 29	1.000 1	0.850 24	0.924 3	0.648 25	0.747 52	0.162 40	0.862 3	0.572 34	0.520 34	0.624 27	0.549 33	0.649 37	1.000 1	0.560 45	0.706 52	0.768 21	1.000 1	0.591 26

HAIS	0.699 30	1.000 1	0.849 25	0.820 22	0.675 23	0.808 38	0.279 30	0.757 16	0.465 41	0.517 35	0.596 32	0.559 32	0.600 44	1.000 1	0.654 33	0.767 37	0.676 33	0.994 50	0.560 32
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
SSTNet	0.698 31	1.000 1	0.697 63	0.888 8	0.556 42	0.803 39	0.387 24	0.626 46	0.417 46	0.556 31	0.585 36	0.702 15	0.600 44	1.000 1	0.824 8	0.720 50	0.692 31	1.000 1	0.509 40
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
DualGroup	0.694 32	1.000 1	0.799 41	0.811 23	0.622 29	0.817 33	0.376 25	0.805 9	0.590 33	0.487 40	0.568 39	0.525 37	0.650 28	0.835 58	0.600 41	0.829 21	0.655 36	1.000 1	0.526 37

ODIN - Ins	0.693 33	1.000 1	0.880 21	0.647 48	0.620 31	0.779 46	0.336 27	0.501 61	0.681 21	0.577 27	0.595 33	0.679 24	0.683 25	1.000 1	0.709 26	0.816 29	0.637 40	0.770 70	0.557 33
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
DANCENET	0.680 34	1.000 1	0.807 36	0.733 38	0.600 36	0.768 48	0.375 26	0.543 54	0.538 36	0.610 24	0.599 31	0.498 38	0.632 42	0.981 40	0.739 23	0.856 9	0.633 43	0.882 65	0.454 50

SphereSeg	0.680 34	1.000 1	0.856 23	0.744 37	0.618 33	0.893 16	0.151 41	0.651 41	0.713 15	0.537 33	0.579 38	0.430 47	0.651 27	1.000 1	0.389 61	0.744 45	0.697 30	0.991 52	0.601 22

Box2Mask	0.677 36	1.000 1	0.847 26	0.771 32	0.509 51	0.816 34	0.277 31	0.558 53	0.482 38	0.562 30	0.640 24	0.448 43	0.700 16	1.000 1	0.666 29	0.852 13	0.578 50	0.997 41	0.488 45
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 37	1.000 1	0.758 55	0.682 45	0.576 40	0.842 29	0.477 18	0.504 60	0.524 37	0.567 29	0.585 37	0.451 42	0.557 52	1.000 1	0.751 22	0.797 32	0.563 53	1.000 1	0.467 49
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
Mask-Group	0.664 38	1.000 1	0.822 30	0.764 35	0.616 34	0.815 35	0.139 45	0.694 29	0.597 32	0.459 44	0.566 40	0.599 31	0.600 44	0.516 68	0.715 25	0.819 27	0.635 41	1.000 1	0.603 20
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 39	1.000 1	0.760 53	0.667 47	0.581 38	0.863 24	0.323 28	0.655 39	0.477 39	0.473 42	0.549 42	0.432 46	0.650 28	1.000 1	0.655 32	0.738 46	0.585 49	0.944 57	0.472 48

CSC-Pretrained	0.648 40	1.000 1	0.810 33	0.768 33	0.523 49	0.813 36	0.143 44	0.819 5	0.389 49	0.422 53	0.511 46	0.443 44	0.650 28	1.000 1	0.624 38	0.732 47	0.634 42	1.000 1	0.375 57

PE	0.645 41	1.000 1	0.773 49	0.798 27	0.538 44	0.786 45	0.088 53	0.799 10	0.350 53	0.435 51	0.547 43	0.545 34	0.646 39	0.933 47	0.562 44	0.761 40	0.556 58	0.997 41	0.501 43
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
RPGN	0.643 42	1.000 1	0.758 54	0.582 59	0.539 43	0.826 32	0.046 58	0.765 12	0.372 51	0.436 50	0.588 34	0.539 36	0.650 28	1.000 1	0.577 42	0.750 43	0.653 38	0.997 41	0.495 44
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
Dyco3D	0.641 43	1.000 1	0.841 27	0.893 7	0.531 46	0.802 40	0.115 50	0.588 51	0.448 43	0.438 48	0.537 45	0.430 48	0.550 53	0.857 50	0.534 48	0.764 39	0.657 35	0.987 53	0.568 29
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
GICN	0.638 44	1.000 1	0.895 19	0.800 26	0.480 55	0.676 57	0.144 43	0.737 18	0.354 52	0.447 45	0.400 59	0.365 54	0.700 16	1.000 1	0.569 43	0.836 18	0.599 45	1.000 1	0.473 47

PointGroup	0.636 45	1.000 1	0.765 50	0.624 50	0.505 53	0.797 41	0.116 49	0.696 28	0.384 50	0.441 46	0.559 41	0.476 40	0.596 47	1.000 1	0.666 29	0.756 41	0.556 57	0.997 41	0.513 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]
DD-UNet+Group	0.635 46	0.667 61	0.797 43	0.714 42	0.562 41	0.774 47	0.146 42	0.810 8	0.429 45	0.476 41	0.546 44	0.399 50	0.633 40	1.000 1	0.632 37	0.722 49	0.609 44	1.000 1	0.514 38
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 47	1.000 1	0.829 29	0.606 52	0.646 26	0.836 30	0.068 54	0.511 58	0.462 42	0.507 37	0.619 29	0.389 52	0.610 43	1.000 1	0.432 56	0.828 23	0.673 34	0.788 69	0.552 34

DENet	0.629 48	1.000 1	0.797 42	0.608 51	0.589 37	0.627 61	0.219 37	0.882 1	0.310 55	0.402 58	0.383 61	0.396 51	0.650 28	1.000 1	0.663 31	0.543 69	0.691 32	1.000 1	0.568 30

3D-MPA	0.611 49	1.000 1	0.833 28	0.765 34	0.526 48	0.756 50	0.136 47	0.588 51	0.470 40	0.438 49	0.432 55	0.358 56	0.650 28	0.857 50	0.429 57	0.765 38	0.557 56	1.000 1	0.430 52
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 50	1.000 1	0.801 40	0.599 54	0.535 45	0.728 54	0.286 29	0.436 65	0.679 22	0.491 38	0.433 53	0.256 58	0.404 65	0.857 50	0.620 39	0.724 48	0.510 63	1.000 1	0.539 36

AOIA	0.601 51	1.000 1	0.761 52	0.687 44	0.485 54	0.828 31	0.008 65	0.663 37	0.405 48	0.405 57	0.425 56	0.490 39	0.596 47	0.714 61	0.553 47	0.779 33	0.597 46	0.992 51	0.424 54

PCJC	0.578 52	1.000 1	0.810 34	0.583 58	0.449 58	0.813 37	0.042 59	0.603 49	0.341 54	0.490 39	0.465 50	0.410 49	0.650 28	0.835 58	0.264 67	0.694 56	0.561 54	0.889 62	0.504 42

SSEN	0.575 53	1.000 1	0.761 51	0.473 61	0.477 56	0.795 42	0.066 55	0.529 56	0.658 27	0.460 43	0.461 51	0.380 53	0.331 67	0.859 49	0.401 60	0.692 58	0.653 37	1.000 1	0.348 59
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 54	0.528 71	0.708 62	0.626 49	0.580 39	0.745 53	0.063 56	0.627 45	0.240 59	0.400 59	0.497 47	0.464 41	0.515 54	1.000 1	0.475 52	0.745 44	0.571 51	1.000 1	0.429 53

NeuralBF	0.555 55	0.667 61	0.896 18	0.843 19	0.517 50	0.751 51	0.029 60	0.519 57	0.414 47	0.439 47	0.465 49	0.000 77	0.484 56	0.857 50	0.287 65	0.693 57	0.651 39	1.000 1	0.485 46
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 56	1.000 1	0.807 37	0.588 57	0.327 63	0.647 59	0.004 67	0.815 7	0.180 62	0.418 54	0.364 63	0.182 61	0.445 59	1.000 1	0.442 55	0.688 59	0.571 52	1.000 1	0.396 55
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
ClickSeg_Instance	0.539 57	1.000 1	0.621 66	0.300 64	0.530 47	0.698 55	0.127 48	0.533 55	0.222 60	0.430 52	0.400 58	0.365 54	0.574 51	0.938 46	0.472 53	0.659 61	0.543 59	0.944 57	0.347 60

One_Thing_One_Click	0.529 58	0.667 61	0.718 58	0.777 31	0.399 59	0.683 56	0.000 70	0.669 35	0.138 65	0.391 60	0.374 62	0.539 35	0.360 66	0.641 65	0.556 46	0.774 36	0.593 47	0.997 41	0.251 65
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 59	1.000 1	0.538 71	0.282 65	0.468 57	0.790 43	0.173 39	0.345 67	0.429 44	0.413 56	0.484 48	0.176 62	0.595 49	0.591 66	0.522 49	0.668 60	0.476 64	0.986 55	0.327 61

Occipital-SCS	0.512 60	1.000 1	0.716 59	0.509 60	0.506 52	0.611 62	0.092 52	0.602 50	0.177 63	0.346 63	0.383 60	0.165 63	0.442 60	0.850 57	0.386 62	0.618 65	0.543 60	0.889 62	0.389 56

3D-BoNet	0.488 61	1.000 1	0.672 65	0.590 56	0.301 65	0.484 72	0.098 51	0.620 47	0.306 56	0.341 64	0.259 67	0.125 65	0.434 62	0.796 60	0.402 59	0.499 71	0.513 62	0.909 61	0.439 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	0.478 62	0.667 61	0.712 61	0.595 55	0.259 68	0.550 68	0.000 70	0.613 48	0.175 64	0.250 69	0.434 52	0.437 45	0.411 64	0.857 50	0.485 51	0.591 68	0.267 74	0.944 57	0.359 58
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 63	0.667 61	0.685 64	0.677 46	0.372 61	0.562 66	0.000 70	0.482 62	0.244 58	0.316 66	0.298 64	0.052 72	0.442 61	0.857 50	0.267 66	0.702 53	0.559 55	1.000 1	0.287 63

SALoss-ResNet	0.459 64	1.000 1	0.737 57	0.159 75	0.259 67	0.587 64	0.138 46	0.475 63	0.217 61	0.416 55	0.408 57	0.128 64	0.315 68	0.714 61	0.411 58	0.536 70	0.590 48	0.873 66	0.304 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)
MASC	0.447 65	0.528 71	0.555 69	0.381 62	0.382 60	0.633 60	0.002 68	0.509 59	0.260 57	0.361 62	0.432 54	0.327 57	0.451 58	0.571 67	0.367 63	0.639 63	0.386 65	0.980 56	0.276 64
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
SegGroup_ins	0.445 66	0.667 61	0.773 48	0.185 72	0.317 64	0.656 58	0.000 70	0.407 66	0.134 66	0.381 61	0.267 66	0.217 60	0.476 57	0.714 61	0.452 54	0.629 64	0.514 61	1.000 1	0.222 68
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 67	1.000 1	0.432 74	0.245 67	0.190 69	0.577 65	0.013 64	0.263 69	0.033 72	0.320 65	0.240 68	0.075 68	0.422 63	0.857 50	0.117 72	0.699 54	0.271 73	0.883 64	0.235 67
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
Hier3D	0.323 68	0.667 61	0.542 70	0.264 66	0.157 72	0.550 67	0.000 70	0.205 72	0.009 74	0.270 68	0.218 69	0.075 68	0.500 55	0.688 64	0.007 78	0.698 55	0.301 70	0.459 75	0.200 69
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
UNet-backbone	0.319 69	0.667 61	0.715 60	0.233 68	0.189 70	0.479 73	0.008 65	0.218 70	0.067 71	0.201 71	0.173 70	0.107 66	0.123 73	0.438 69	0.150 69	0.615 66	0.355 66	0.916 60	0.093 77

R-PointNet	0.306 70	0.500 73	0.405 75	0.311 63	0.348 62	0.589 63	0.054 57	0.068 75	0.126 67	0.283 67	0.290 65	0.028 73	0.219 71	0.214 72	0.331 64	0.396 75	0.275 71	0.821 68	0.245 66

Region-18class	0.284 71	0.250 77	0.751 56	0.228 70	0.270 66	0.521 69	0.000 70	0.468 64	0.008 76	0.205 70	0.127 71	0.000 77	0.068 75	0.070 76	0.262 68	0.652 62	0.323 68	0.740 71	0.173 70

SemRegionNet-20cls	0.250 72	0.333 74	0.613 67	0.229 69	0.163 71	0.493 70	0.000 70	0.304 68	0.107 68	0.147 74	0.100 73	0.052 71	0.231 69	0.119 74	0.039 74	0.445 73	0.325 67	0.654 72	0.141 73

tmp	0.248 73	0.667 61	0.437 73	0.188 71	0.153 73	0.491 71	0.000 70	0.208 71	0.094 70	0.153 73	0.099 74	0.057 70	0.217 72	0.119 74	0.039 74	0.466 72	0.302 69	0.640 73	0.140 74

3D-BEVIS	0.248 73	0.667 61	0.566 68	0.076 76	0.035 78	0.394 76	0.027 62	0.035 77	0.098 69	0.099 76	0.030 77	0.025 74	0.098 74	0.375 71	0.126 71	0.604 67	0.181 76	0.854 67	0.171 71
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
Sem_Recon_ins	0.227 75	0.764 60	0.486 72	0.069 77	0.098 75	0.426 75	0.017 63	0.067 76	0.015 73	0.172 72	0.100 72	0.096 67	0.054 77	0.183 73	0.135 70	0.366 76	0.260 75	0.614 74	0.168 72

ASIS	0.199 76	0.333 74	0.253 77	0.167 74	0.140 74	0.438 74	0.000 70	0.177 73	0.008 75	0.121 75	0.069 75	0.004 76	0.231 70	0.429 70	0.036 76	0.445 74	0.273 72	0.333 77	0.119 76

Sgpn_scannet	0.143 77	0.208 78	0.390 76	0.169 73	0.065 76	0.275 77	0.029 61	0.069 74	0.000 77	0.087 77	0.043 76	0.014 75	0.027 78	0.000 77	0.112 73	0.351 77	0.168 77	0.438 76	0.138 75

MaskRCNN 2d->3d Proj	0.058 78	0.333 74	0.002 78	0.000 78	0.053 77	0.002 78	0.002 69	0.021 78	0.000 77	0.045 78	0.024 78	0.238 59	0.065 76	0.000 77	0.014 77	0.107 78	0.020 78	0.110 78	0.006 78

3D Semantic Instance Benchmark