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

MAFT	0.786 14	1.000 1	0.894 19	0.807 23	0.694 17	0.893 15	0.486 15	0.674 31	0.740 13	0.786 1	0.704 11	0.727 10	0.739 14	1.000 1	0.707 24	0.849 13	0.756 24	1.000 1	0.685 1

OneFormer3D	0.801 6	1.000 1	0.973 6	0.909 5	0.698 15	0.928 4	0.582 4	0.668 34	0.685 19	0.780 2	0.687 14	0.698 19	0.702 15	1.000 1	0.794 12	0.900 2	0.784 13	0.986 52	0.635 8
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
Competitor-MAFT	0.816 1	1.000 1	0.983 3	0.872 10	0.718 5	0.941 1	0.588 3	0.652 38	0.819 2	0.776 3	0.720 5	0.780 4	0.769 12	1.000 1	0.797 10	0.813 27	0.798 7	1.000 1	0.659 3

InsSSM	0.799 9	1.000 1	0.915 14	0.710 41	0.729 3	0.925 5	0.664 1	0.670 32	0.770 7	0.766 4	0.739 3	0.737 7	0.700 16	1.000 1	0.792 13	0.829 20	0.815 4	0.997 39	0.625 10
Lei Yao, Yi Wang, Moyun Liu, Lap-Pui Chau: SGIFormer: Semantic-guided and Geometric-enhanced Interleaving Transformer for 3D Instance Segmentation. TCSVT, 2024
MG-Former	0.791 11	1.000 1	0.980 5	0.837 19	0.626 26	0.897 12	0.543 12	0.759 15	0.800 6	0.766 5	0.659 17	0.769 5	0.697 19	1.000 1	0.791 14	0.707 48	0.791 10	1.000 1	0.610 16

ExtMask3D	0.789 12	1.000 1	0.988 1	0.756 34	0.706 12	0.912 11	0.429 20	0.647 40	0.806 5	0.755 6	0.673 15	0.689 20	0.772 11	1.000 1	0.789 15	0.852 11	0.811 5	1.000 1	0.617 13

UniPerception	0.800 7	1.000 1	0.930 12	0.872 10	0.727 4	0.862 24	0.454 19	0.764 13	0.820 1	0.746 7	0.706 10	0.750 6	0.772 10	0.926 45	0.764 18	0.818 26	0.826 1	0.997 39	0.660 2

TST3D	0.795 10	1.000 1	0.929 13	0.918 4	0.709 10	0.884 19	0.596 2	0.704 25	0.769 8	0.734 8	0.644 21	0.699 17	0.751 13	1.000 1	0.794 11	0.876 5	0.757 23	0.997 39	0.550 32
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
PointRel	0.816 1	1.000 1	0.971 8	0.908 6	0.743 2	0.923 7	0.573 7	0.714 22	0.695 18	0.734 9	0.747 2	0.725 11	0.809 1	1.000 1	0.814 8	0.899 3	0.820 3	1.000 1	0.610 17
: Relation3D (PointRel): Enhancing Relation Modeling for Point Cloud Instance Segmentation.
Competitor-SPFormer	0.800 7	1.000 1	0.986 2	0.845 16	0.705 13	0.915 10	0.532 13	0.733 20	0.757 10	0.733 10	0.708 9	0.698 18	0.648 35	0.981 38	0.890 1	0.830 18	0.796 8	0.997 39	0.644 4

SIM3D	0.803 5	1.000 1	0.967 10	0.863 13	0.692 18	0.924 6	0.552 11	0.732 21	0.667 22	0.732 11	0.662 16	0.796 1	0.789 9	1.000 1	0.803 9	0.864 6	0.766 20	1.000 1	0.643 5

EV3D	0.811 4	1.000 1	0.968 9	0.852 14	0.717 7	0.921 8	0.574 6	0.677 29	0.748 11	0.730 12	0.703 12	0.795 2	0.809 1	1.000 1	0.831 4	0.854 9	0.778 15	1.000 1	0.638 7

Spherical Mask(CtoF)	0.812 3	1.000 1	0.973 7	0.852 14	0.718 6	0.917 9	0.574 5	0.677 29	0.748 11	0.729 13	0.715 7	0.795 2	0.809 1	1.000 1	0.831 4	0.854 9	0.787 11	1.000 1	0.638 6

KmaxOneFormerNet	0.783 15	0.903 55	0.981 4	0.794 27	0.706 11	0.931 3	0.561 9	0.701 26	0.706 16	0.727 14	0.697 13	0.731 9	0.689 22	1.000 1	0.856 2	0.750 39	0.761 22	1.000 1	0.599 21

SPFormer	0.770 17	0.903 55	0.903 16	0.806 24	0.609 32	0.886 16	0.568 8	0.815 6	0.705 17	0.711 15	0.655 18	0.652 24	0.685 23	1.000 1	0.789 16	0.809 28	0.776 17	1.000 1	0.583 25
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
ISBNet	0.757 20	1.000 1	0.904 15	0.731 37	0.678 20	0.895 13	0.458 17	0.644 42	0.670 21	0.710 16	0.620 26	0.732 8	0.650 25	1.000 1	0.756 19	0.778 31	0.779 14	1.000 1	0.614 14
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
DKNet	0.718 26	1.000 1	0.814 29	0.782 28	0.619 29	0.872 20	0.224 33	0.751 17	0.569 32	0.677 17	0.585 32	0.724 12	0.633 37	0.981 38	0.515 47	0.819 24	0.736 27	1.000 1	0.617 12
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
Mask3D	0.780 16	1.000 1	0.786 43	0.716 39	0.696 16	0.885 18	0.500 14	0.714 22	0.810 4	0.672 18	0.715 7	0.679 21	0.809 1	1.000 1	0.831 4	0.833 17	0.787 11	1.000 1	0.602 19
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
Queryformer	0.787 13	1.000 1	0.933 11	0.601 50	0.754 1	0.886 17	0.558 10	0.661 36	0.767 9	0.665 19	0.716 6	0.639 25	0.808 5	1.000 1	0.844 3	0.897 4	0.804 6	1.000 1	0.624 11

SoftGroup	0.761 19	1.000 1	0.808 32	0.845 16	0.716 8	0.862 23	0.243 32	0.824 4	0.655 25	0.620 20	0.734 4	0.699 16	0.791 8	0.981 38	0.716 22	0.844 14	0.769 18	1.000 1	0.594 23
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral]
PBNet	0.747 22	1.000 1	0.818 28	0.837 20	0.713 9	0.844 26	0.457 18	0.647 40	0.711 15	0.614 21	0.617 28	0.657 23	0.650 25	1.000 1	0.692 25	0.822 22	0.765 21	1.000 1	0.595 22
W.Zhao, Y.Yan, C.Yang, J.Ye,X.Yang,K.Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
DANCENET	0.680 31	1.000 1	0.807 33	0.733 36	0.600 33	0.768 45	0.375 24	0.543 52	0.538 33	0.610 22	0.599 29	0.498 35	0.632 39	0.981 38	0.739 21	0.856 8	0.633 40	0.882 63	0.454 47

TD3D	0.751 21	1.000 1	0.774 44	0.867 12	0.621 28	0.934 2	0.404 21	0.706 24	0.812 3	0.605 23	0.633 24	0.626 26	0.690 21	1.000 1	0.640 33	0.820 23	0.777 16	1.000 1	0.612 15
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
GraphCut	0.732 23	1.000 1	0.788 41	0.724 38	0.642 25	0.859 25	0.248 31	0.787 11	0.618 28	0.596 24	0.653 20	0.722 13	0.583 47	1.000 1	0.766 17	0.861 7	0.825 2	1.000 1	0.504 38

SoftGroup++	0.769 18	1.000 1	0.803 36	0.937 1	0.684 19	0.865 21	0.213 35	0.870 2	0.664 23	0.571 25	0.758 1	0.702 15	0.807 6	1.000 1	0.653 31	0.902 1	0.792 9	1.000 1	0.626 9

OccuSeg+instance	0.672 34	1.000 1	0.758 52	0.682 43	0.576 37	0.842 27	0.477 16	0.504 58	0.524 34	0.567 26	0.585 34	0.451 39	0.557 49	1.000 1	0.751 20	0.797 29	0.563 50	1.000 1	0.467 46
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
Box2Mask	0.677 33	1.000 1	0.847 23	0.771 30	0.509 48	0.816 32	0.277 28	0.558 51	0.482 35	0.562 27	0.640 22	0.448 40	0.700 16	1.000 1	0.666 26	0.852 12	0.578 47	0.997 39	0.488 42
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022
SSTNet	0.698 29	1.000 1	0.697 60	0.888 8	0.556 39	0.803 37	0.387 22	0.626 44	0.417 43	0.556 28	0.585 33	0.702 14	0.600 41	1.000 1	0.824 7	0.720 47	0.692 29	1.000 1	0.509 37
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
TopoSeg	0.725 25	1.000 1	0.806 35	0.933 2	0.668 22	0.758 46	0.272 29	0.734 19	0.630 27	0.549 29	0.654 19	0.606 27	0.697 20	0.966 42	0.612 37	0.839 15	0.754 25	1.000 1	0.573 26

SphereSeg	0.680 31	1.000 1	0.856 20	0.744 35	0.618 30	0.893 14	0.151 38	0.651 39	0.713 14	0.537 30	0.579 35	0.430 44	0.651 24	1.000 1	0.389 58	0.744 42	0.697 28	0.991 50	0.601 20

SSEC	0.707 27	1.000 1	0.850 21	0.924 3	0.648 23	0.747 49	0.162 37	0.862 3	0.572 31	0.520 31	0.624 25	0.549 30	0.649 34	1.000 1	0.560 42	0.706 49	0.768 19	1.000 1	0.591 24

HAIS	0.699 28	1.000 1	0.849 22	0.820 21	0.675 21	0.808 36	0.279 27	0.757 16	0.465 38	0.517 32	0.596 30	0.559 29	0.600 41	1.000 1	0.654 30	0.767 34	0.676 31	0.994 48	0.560 30
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
IPCA-Inst	0.731 24	1.000 1	0.788 42	0.884 9	0.698 14	0.788 42	0.252 30	0.760 14	0.646 26	0.511 33	0.637 23	0.665 22	0.804 7	1.000 1	0.644 32	0.778 32	0.747 26	1.000 1	0.561 29

Mask3D_evaluation	0.631 44	1.000 1	0.829 26	0.606 49	0.646 24	0.836 28	0.068 51	0.511 56	0.462 39	0.507 34	0.619 27	0.389 49	0.610 40	1.000 1	0.432 53	0.828 21	0.673 32	0.788 67	0.552 31

OSIS	0.605 47	1.000 1	0.801 37	0.599 51	0.535 42	0.728 51	0.286 26	0.436 62	0.679 20	0.491 35	0.433 50	0.256 55	0.404 62	0.857 47	0.620 36	0.724 45	0.510 60	1.000 1	0.539 33

PCJC	0.578 49	1.000 1	0.810 31	0.583 55	0.449 55	0.813 35	0.042 56	0.603 47	0.341 51	0.490 36	0.465 47	0.410 46	0.650 25	0.835 55	0.264 64	0.694 53	0.561 51	0.889 60	0.504 39

DualGroup	0.694 30	1.000 1	0.799 38	0.811 22	0.622 27	0.817 31	0.376 23	0.805 9	0.590 30	0.487 37	0.568 36	0.525 34	0.650 25	0.835 55	0.600 38	0.829 19	0.655 34	1.000 1	0.526 34

DD-UNet+Group	0.635 43	0.667 58	0.797 40	0.714 40	0.562 38	0.774 44	0.146 39	0.810 8	0.429 42	0.476 38	0.546 41	0.399 47	0.633 37	1.000 1	0.632 34	0.722 46	0.609 41	1.000 1	0.514 35
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	0.657 36	1.000 1	0.760 50	0.667 45	0.581 35	0.863 22	0.323 25	0.655 37	0.477 36	0.473 39	0.549 39	0.432 43	0.650 25	1.000 1	0.655 29	0.738 43	0.585 46	0.944 55	0.472 45

SSEN	0.575 50	1.000 1	0.761 48	0.473 58	0.477 53	0.795 40	0.066 52	0.529 54	0.658 24	0.460 40	0.461 48	0.380 50	0.331 64	0.859 46	0.401 57	0.692 55	0.653 35	1.000 1	0.348 56
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv
Mask-Group	0.664 35	1.000 1	0.822 27	0.764 33	0.616 31	0.815 33	0.139 42	0.694 28	0.597 29	0.459 41	0.566 37	0.599 28	0.600 41	0.516 65	0.715 23	0.819 25	0.635 38	1.000 1	0.603 18
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022
GICN	0.638 41	1.000 1	0.895 18	0.800 25	0.480 52	0.676 54	0.144 40	0.737 18	0.354 49	0.447 42	0.400 56	0.365 51	0.700 16	1.000 1	0.569 40	0.836 16	0.599 42	1.000 1	0.473 44

PointGroup	0.636 42	1.000 1	0.765 47	0.624 47	0.505 50	0.797 39	0.116 46	0.696 27	0.384 47	0.441 43	0.559 38	0.476 37	0.596 44	1.000 1	0.666 26	0.756 38	0.556 54	0.997 39	0.513 36
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]
NeuralBF	0.555 52	0.667 58	0.896 17	0.843 18	0.517 47	0.751 48	0.029 57	0.519 55	0.414 44	0.439 44	0.465 46	0.000 74	0.484 53	0.857 47	0.287 62	0.693 54	0.651 37	1.000 1	0.485 43
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
Dyco3D	0.641 40	1.000 1	0.841 24	0.893 7	0.531 43	0.802 38	0.115 47	0.588 49	0.448 40	0.438 45	0.537 42	0.430 45	0.550 50	0.857 47	0.534 45	0.764 36	0.657 33	0.987 51	0.568 27
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
3D-MPA	0.611 46	1.000 1	0.833 25	0.765 32	0.526 45	0.756 47	0.136 44	0.588 49	0.470 37	0.438 46	0.432 52	0.358 53	0.650 25	0.857 47	0.429 54	0.765 35	0.557 53	1.000 1	0.430 49
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020
RPGN	0.643 39	1.000 1	0.758 51	0.582 56	0.539 40	0.826 30	0.046 55	0.765 12	0.372 48	0.436 47	0.588 31	0.539 33	0.650 25	1.000 1	0.577 39	0.750 40	0.653 36	0.997 39	0.495 41
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
PE	0.645 38	1.000 1	0.773 46	0.798 26	0.538 41	0.786 43	0.088 50	0.799 10	0.350 50	0.435 48	0.547 40	0.545 31	0.646 36	0.933 44	0.562 41	0.761 37	0.556 55	0.997 39	0.501 40
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
ClickSeg_Instance	0.539 54	1.000 1	0.621 63	0.300 61	0.530 44	0.698 52	0.127 45	0.533 53	0.222 57	0.430 49	0.400 55	0.365 51	0.574 48	0.938 43	0.472 50	0.659 58	0.543 56	0.944 55	0.347 57

CSC-Pretrained	0.648 37	1.000 1	0.810 30	0.768 31	0.523 46	0.813 34	0.143 41	0.819 5	0.389 46	0.422 50	0.511 43	0.443 41	0.650 25	1.000 1	0.624 35	0.732 44	0.634 39	1.000 1	0.375 54

MTML	0.549 53	1.000 1	0.807 34	0.588 54	0.327 60	0.647 56	0.004 64	0.815 7	0.180 59	0.418 51	0.364 60	0.182 58	0.445 56	1.000 1	0.442 52	0.688 56	0.571 49	1.000 1	0.396 52
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
SALoss-ResNet	0.459 61	1.000 1	0.737 54	0.159 72	0.259 64	0.587 61	0.138 43	0.475 60	0.217 58	0.416 52	0.408 54	0.128 61	0.315 65	0.714 58	0.411 55	0.536 67	0.590 45	0.873 64	0.304 59
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)
Sparse R-CNN	0.515 56	1.000 1	0.538 68	0.282 62	0.468 54	0.790 41	0.173 36	0.345 64	0.429 41	0.413 53	0.484 45	0.176 59	0.595 46	0.591 63	0.522 46	0.668 57	0.476 61	0.986 53	0.327 58

AOIA	0.601 48	1.000 1	0.761 49	0.687 42	0.485 51	0.828 29	0.008 62	0.663 35	0.405 45	0.405 54	0.425 53	0.490 36	0.596 44	0.714 58	0.553 44	0.779 30	0.597 43	0.992 49	0.424 51

DENet	0.629 45	1.000 1	0.797 39	0.608 48	0.589 34	0.627 58	0.219 34	0.882 1	0.310 52	0.402 55	0.383 58	0.396 48	0.650 25	1.000 1	0.663 28	0.543 66	0.691 30	1.000 1	0.568 28

RWSeg	0.567 51	0.528 68	0.708 59	0.626 46	0.580 36	0.745 50	0.063 53	0.627 43	0.240 56	0.400 56	0.497 44	0.464 38	0.515 51	1.000 1	0.475 49	0.745 41	0.571 48	1.000 1	0.429 50

One_Thing_One_Click	0.529 55	0.667 58	0.718 55	0.777 29	0.399 56	0.683 53	0.000 67	0.669 33	0.138 62	0.391 57	0.374 59	0.539 32	0.360 63	0.641 62	0.556 43	0.774 33	0.593 44	0.997 39	0.251 62
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
SegGroup_ins	0.445 63	0.667 58	0.773 45	0.185 69	0.317 61	0.656 55	0.000 67	0.407 63	0.134 63	0.381 58	0.267 63	0.217 57	0.476 54	0.714 58	0.452 51	0.629 61	0.514 58	1.000 1	0.222 65
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
MASC	0.447 62	0.528 68	0.555 66	0.381 59	0.382 57	0.633 57	0.002 65	0.509 57	0.260 54	0.361 59	0.432 51	0.327 54	0.451 55	0.571 64	0.367 60	0.639 60	0.386 62	0.980 54	0.276 61
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
Occipital-SCS	0.512 57	1.000 1	0.716 56	0.509 57	0.506 49	0.611 59	0.092 49	0.602 48	0.177 60	0.346 60	0.383 57	0.165 60	0.442 57	0.850 54	0.386 59	0.618 62	0.543 57	0.889 60	0.389 53

3D-BoNet	0.488 58	1.000 1	0.672 62	0.590 53	0.301 62	0.484 69	0.098 48	0.620 45	0.306 53	0.341 61	0.259 64	0.125 62	0.434 59	0.796 57	0.402 56	0.499 68	0.513 59	0.909 59	0.439 48
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
3D-SIS	0.382 64	1.000 1	0.432 71	0.245 64	0.190 66	0.577 62	0.013 61	0.263 66	0.033 69	0.320 62	0.240 65	0.075 65	0.422 60	0.857 47	0.117 69	0.699 51	0.271 70	0.883 62	0.235 64
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
SPG_WSIS	0.470 60	0.667 58	0.685 61	0.677 44	0.372 58	0.562 63	0.000 67	0.482 59	0.244 55	0.316 63	0.298 61	0.052 69	0.442 58	0.857 47	0.267 63	0.702 50	0.559 52	1.000 1	0.287 60

R-PointNet	0.306 67	0.500 70	0.405 72	0.311 60	0.348 59	0.589 60	0.054 54	0.068 72	0.126 64	0.283 64	0.290 62	0.028 70	0.219 68	0.214 69	0.331 61	0.396 72	0.275 68	0.821 66	0.245 63

Hier3D	0.323 65	0.667 58	0.542 67	0.264 63	0.157 69	0.550 64	0.000 67	0.205 69	0.009 71	0.270 65	0.218 66	0.075 65	0.500 52	0.688 61	0.007 75	0.698 52	0.301 67	0.459 72	0.200 66
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
PanopticFusion-inst	0.478 59	0.667 58	0.712 58	0.595 52	0.259 65	0.550 65	0.000 67	0.613 46	0.175 61	0.250 66	0.434 49	0.437 42	0.411 61	0.857 47	0.485 48	0.591 65	0.267 71	0.944 55	0.359 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)
Region-18class	0.284 68	0.250 74	0.751 53	0.228 67	0.270 63	0.521 66	0.000 67	0.468 61	0.008 73	0.205 67	0.127 68	0.000 74	0.068 72	0.070 73	0.262 65	0.652 59	0.323 65	0.740 68	0.173 67

UNet-backbone	0.319 66	0.667 58	0.715 57	0.233 65	0.189 67	0.479 70	0.008 62	0.218 67	0.067 68	0.201 68	0.173 67	0.107 63	0.123 70	0.438 66	0.150 66	0.615 63	0.355 63	0.916 58	0.093 74

Sem_Recon_ins	0.227 72	0.764 57	0.486 69	0.069 74	0.098 72	0.426 72	0.017 60	0.067 73	0.015 70	0.172 69	0.100 69	0.096 64	0.054 74	0.183 70	0.135 67	0.366 73	0.260 72	0.614 71	0.168 69

tmp	0.248 70	0.667 58	0.437 70	0.188 68	0.153 70	0.491 68	0.000 67	0.208 68	0.094 67	0.153 70	0.099 71	0.057 67	0.217 69	0.119 71	0.039 71	0.466 69	0.302 66	0.640 70	0.140 71

SemRegionNet-20cls	0.250 69	0.333 71	0.613 64	0.229 66	0.163 68	0.493 67	0.000 67	0.304 65	0.107 65	0.147 71	0.100 70	0.052 68	0.231 66	0.119 71	0.039 71	0.445 70	0.325 64	0.654 69	0.141 70

ASIS	0.199 73	0.333 71	0.253 74	0.167 71	0.140 71	0.438 71	0.000 67	0.177 70	0.008 72	0.121 72	0.069 72	0.004 73	0.231 67	0.429 67	0.036 73	0.445 71	0.273 69	0.333 74	0.119 73

3D-BEVIS	0.248 70	0.667 58	0.566 65	0.076 73	0.035 75	0.394 73	0.027 59	0.035 74	0.098 66	0.099 73	0.030 74	0.025 71	0.098 71	0.375 68	0.126 68	0.604 64	0.181 73	0.854 65	0.171 68
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
Sgpn_scannet	0.143 74	0.208 75	0.390 73	0.169 70	0.065 73	0.275 74	0.029 58	0.069 71	0.000 74	0.087 74	0.043 73	0.014 72	0.027 75	0.000 74	0.112 70	0.351 74	0.168 74	0.438 73	0.138 72

MaskRCNN 2d->3d Proj	0.058 75	0.333 71	0.002 75	0.000 75	0.053 74	0.002 75	0.002 66	0.021 75	0.000 74	0.045 75	0.024 75	0.238 56	0.065 73	0.000 74	0.014 74	0.107 75	0.020 75	0.110 75	0.006 75

3D Semantic Instance Benchmark