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

PointRel	0.816 1	1.000 1	0.971 6	0.908 6	0.743 2	0.923 5	0.573 6	0.714 22	0.695 16	0.734 8	0.747 2	0.725 9	0.809 1	1.000 1	0.814 7	0.899 3	0.820 3	1.000 1	0.610 16
: Relation3D (PointRel): Enhancing Relation Modeling for Point Cloud Instance Segmentation.
Spherical Mask(CtoF)	0.812 2	1.000 1	0.973 5	0.852 13	0.718 5	0.917 7	0.574 4	0.677 28	0.748 10	0.729 12	0.715 6	0.795 2	0.809 1	1.000 1	0.831 3	0.854 9	0.787 10	1.000 1	0.638 5

EV3D	0.811 3	1.000 1	0.968 7	0.852 13	0.717 6	0.921 6	0.574 5	0.677 28	0.748 10	0.730 11	0.703 11	0.795 2	0.809 1	1.000 1	0.831 3	0.854 9	0.778 14	1.000 1	0.638 6

SIM3D	0.803 4	1.000 1	0.967 8	0.863 12	0.692 16	0.924 4	0.552 9	0.732 21	0.667 20	0.732 10	0.662 14	0.796 1	0.789 9	1.000 1	0.803 8	0.864 6	0.766 19	1.000 1	0.643 4

OneFormer3D	0.801 5	1.000 1	0.973 4	0.909 5	0.698 13	0.928 2	0.582 3	0.668 33	0.685 17	0.780 2	0.687 12	0.698 17	0.702 14	1.000 1	0.794 10	0.900 2	0.784 12	0.986 50	0.635 7
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
Competitor-SPFormer	0.800 6	1.000 1	0.986 2	0.845 15	0.705 11	0.915 8	0.532 11	0.733 20	0.757 9	0.733 9	0.708 8	0.698 16	0.648 33	0.981 36	0.890 1	0.830 18	0.796 7	0.997 37	0.644 3

UniPerception	0.800 6	1.000 1	0.930 10	0.872 10	0.727 4	0.862 22	0.454 17	0.764 13	0.820 1	0.746 6	0.706 9	0.750 5	0.772 10	0.926 43	0.764 16	0.818 26	0.826 1	0.997 37	0.660 2

InsSSM	0.799 8	1.000 1	0.915 12	0.710 39	0.729 3	0.925 3	0.664 1	0.670 31	0.770 6	0.766 3	0.739 3	0.737 6	0.700 15	1.000 1	0.792 11	0.829 20	0.815 4	0.997 37	0.625 9

TST3D	0.795 9	1.000 1	0.929 11	0.918 4	0.709 9	0.884 17	0.596 2	0.704 25	0.769 7	0.734 7	0.644 19	0.699 15	0.751 12	1.000 1	0.794 9	0.876 5	0.757 21	0.997 37	0.550 30
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
MG-Former	0.791 10	1.000 1	0.980 3	0.837 18	0.626 24	0.897 10	0.543 10	0.759 15	0.800 5	0.766 4	0.659 15	0.769 4	0.697 18	1.000 1	0.791 12	0.707 46	0.791 9	1.000 1	0.610 15

ExtMask3D	0.789 11	1.000 1	0.988 1	0.756 32	0.706 10	0.912 9	0.429 18	0.647 38	0.806 4	0.755 5	0.673 13	0.689 18	0.772 11	1.000 1	0.789 13	0.852 11	0.811 5	1.000 1	0.617 12

Queryformer	0.787 12	1.000 1	0.933 9	0.601 48	0.754 1	0.886 15	0.558 8	0.661 35	0.767 8	0.665 17	0.716 5	0.639 23	0.808 5	1.000 1	0.844 2	0.897 4	0.804 6	1.000 1	0.624 10

MAFT	0.786 13	1.000 1	0.894 17	0.807 22	0.694 15	0.893 13	0.486 13	0.674 30	0.740 12	0.786 1	0.704 10	0.727 8	0.739 13	1.000 1	0.707 22	0.849 13	0.756 22	1.000 1	0.685 1

Mask3D	0.780 14	1.000 1	0.786 41	0.716 37	0.696 14	0.885 16	0.500 12	0.714 22	0.810 3	0.672 16	0.715 6	0.679 19	0.809 1	1.000 1	0.831 3	0.833 17	0.787 10	1.000 1	0.602 18
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
SPFormer	0.770 15	0.903 54	0.903 14	0.806 23	0.609 30	0.886 14	0.568 7	0.815 6	0.705 15	0.711 13	0.655 16	0.652 22	0.685 21	1.000 1	0.789 14	0.809 27	0.776 16	1.000 1	0.583 23
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
SoftGroup++	0.769 16	1.000 1	0.803 34	0.937 1	0.684 17	0.865 19	0.213 33	0.870 2	0.664 21	0.571 23	0.758 1	0.702 13	0.807 6	1.000 1	0.653 29	0.902 1	0.792 8	1.000 1	0.626 8

SoftGroup	0.761 17	1.000 1	0.808 30	0.845 15	0.716 7	0.862 21	0.243 30	0.824 4	0.655 23	0.620 18	0.734 4	0.699 14	0.791 8	0.981 36	0.716 20	0.844 14	0.769 17	1.000 1	0.594 21
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 18	1.000 1	0.904 13	0.731 35	0.678 18	0.895 11	0.458 15	0.644 40	0.670 19	0.710 14	0.620 24	0.732 7	0.650 23	1.000 1	0.756 17	0.778 30	0.779 13	1.000 1	0.614 13
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 19	1.000 1	0.774 42	0.867 11	0.621 26	0.934 1	0.404 19	0.706 24	0.812 2	0.605 21	0.633 22	0.626 24	0.690 20	1.000 1	0.640 31	0.820 23	0.777 15	1.000 1	0.612 14
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
PBNet	0.747 20	1.000 1	0.818 26	0.837 19	0.713 8	0.844 24	0.457 16	0.647 38	0.711 14	0.614 19	0.617 26	0.657 21	0.650 23	1.000 1	0.692 23	0.822 22	0.765 20	1.000 1	0.595 20
W.Zhao, Y.Yan, C.Yang, J.Ye,X.Yang,K.Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
GraphCut	0.732 21	1.000 1	0.788 39	0.724 36	0.642 23	0.859 23	0.248 29	0.787 11	0.618 26	0.596 22	0.653 18	0.722 11	0.583 45	1.000 1	0.766 15	0.861 7	0.825 2	1.000 1	0.504 36

IPCA-Inst	0.731 22	1.000 1	0.788 40	0.884 9	0.698 12	0.788 40	0.252 28	0.760 14	0.646 24	0.511 31	0.637 21	0.665 20	0.804 7	1.000 1	0.644 30	0.778 31	0.747 24	1.000 1	0.561 27

TopoSeg	0.725 23	1.000 1	0.806 33	0.933 2	0.668 20	0.758 44	0.272 27	0.734 19	0.630 25	0.549 27	0.654 17	0.606 25	0.697 19	0.966 40	0.612 35	0.839 15	0.754 23	1.000 1	0.573 24

DKNet	0.718 24	1.000 1	0.814 27	0.782 26	0.619 27	0.872 18	0.224 31	0.751 17	0.569 30	0.677 15	0.585 30	0.724 10	0.633 35	0.981 36	0.515 45	0.819 24	0.736 25	1.000 1	0.617 11
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
SSEC	0.707 25	1.000 1	0.850 19	0.924 3	0.648 21	0.747 47	0.162 35	0.862 3	0.572 29	0.520 29	0.624 23	0.549 28	0.649 32	1.000 1	0.560 40	0.706 47	0.768 18	1.000 1	0.591 22

HAIS	0.699 26	1.000 1	0.849 20	0.820 20	0.675 19	0.808 34	0.279 25	0.757 16	0.465 36	0.517 30	0.596 28	0.559 27	0.600 39	1.000 1	0.654 28	0.767 33	0.676 29	0.994 46	0.560 28
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
SSTNet	0.698 27	1.000 1	0.697 58	0.888 8	0.556 37	0.803 35	0.387 20	0.626 42	0.417 41	0.556 26	0.585 31	0.702 12	0.600 39	1.000 1	0.824 6	0.720 45	0.692 27	1.000 1	0.509 35
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
DualGroup	0.694 28	1.000 1	0.799 36	0.811 21	0.622 25	0.817 29	0.376 21	0.805 9	0.590 28	0.487 35	0.568 34	0.525 32	0.650 23	0.835 53	0.600 36	0.829 19	0.655 32	1.000 1	0.526 32

SphereSeg	0.680 29	1.000 1	0.856 18	0.744 33	0.618 28	0.893 12	0.151 36	0.651 37	0.713 13	0.537 28	0.579 33	0.430 42	0.651 22	1.000 1	0.389 56	0.744 40	0.697 26	0.991 48	0.601 19

DANCENET	0.680 29	1.000 1	0.807 31	0.733 34	0.600 31	0.768 43	0.375 22	0.543 50	0.538 31	0.610 20	0.599 27	0.498 33	0.632 37	0.981 36	0.739 19	0.856 8	0.633 38	0.882 61	0.454 45

Box2Mask	0.677 31	1.000 1	0.847 21	0.771 28	0.509 46	0.816 30	0.277 26	0.558 49	0.482 33	0.562 25	0.640 20	0.448 38	0.700 15	1.000 1	0.666 24	0.852 12	0.578 45	0.997 37	0.488 40
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 32	1.000 1	0.758 50	0.682 41	0.576 35	0.842 25	0.477 14	0.504 56	0.524 32	0.567 24	0.585 32	0.451 37	0.557 47	1.000 1	0.751 18	0.797 28	0.563 48	1.000 1	0.467 44
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
Mask-Group	0.664 33	1.000 1	0.822 25	0.764 31	0.616 29	0.815 31	0.139 40	0.694 27	0.597 27	0.459 39	0.566 35	0.599 26	0.600 39	0.516 63	0.715 21	0.819 25	0.635 36	1.000 1	0.603 17
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 34	1.000 1	0.760 48	0.667 43	0.581 33	0.863 20	0.323 23	0.655 36	0.477 34	0.473 37	0.549 37	0.432 41	0.650 23	1.000 1	0.655 27	0.738 41	0.585 44	0.944 53	0.472 43

CSC-Pretrained	0.648 35	1.000 1	0.810 28	0.768 29	0.523 44	0.813 32	0.143 39	0.819 5	0.389 44	0.422 48	0.511 41	0.443 39	0.650 23	1.000 1	0.624 33	0.732 42	0.634 37	1.000 1	0.375 52

PE	0.645 36	1.000 1	0.773 44	0.798 25	0.538 39	0.786 41	0.088 48	0.799 10	0.350 48	0.435 46	0.547 38	0.545 29	0.646 34	0.933 42	0.562 39	0.761 36	0.556 53	0.997 37	0.501 38
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
RPGN	0.643 37	1.000 1	0.758 49	0.582 54	0.539 38	0.826 28	0.046 53	0.765 12	0.372 46	0.436 45	0.588 29	0.539 31	0.650 23	1.000 1	0.577 37	0.750 38	0.653 34	0.997 37	0.495 39
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
Dyco3D	0.641 38	1.000 1	0.841 22	0.893 7	0.531 41	0.802 36	0.115 45	0.588 47	0.448 38	0.438 43	0.537 40	0.430 43	0.550 48	0.857 45	0.534 43	0.764 35	0.657 31	0.987 49	0.568 25
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
GICN	0.638 39	1.000 1	0.895 16	0.800 24	0.480 50	0.676 52	0.144 38	0.737 18	0.354 47	0.447 40	0.400 54	0.365 49	0.700 15	1.000 1	0.569 38	0.836 16	0.599 40	1.000 1	0.473 42

PointGroup	0.636 40	1.000 1	0.765 45	0.624 45	0.505 48	0.797 37	0.116 44	0.696 26	0.384 45	0.441 41	0.559 36	0.476 35	0.596 42	1.000 1	0.666 24	0.756 37	0.556 52	0.997 37	0.513 34
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 41	0.667 56	0.797 38	0.714 38	0.562 36	0.774 42	0.146 37	0.810 8	0.429 40	0.476 36	0.546 39	0.399 45	0.633 35	1.000 1	0.632 32	0.722 44	0.609 39	1.000 1	0.514 33
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 42	1.000 1	0.829 24	0.606 47	0.646 22	0.836 26	0.068 49	0.511 54	0.462 37	0.507 32	0.619 25	0.389 47	0.610 38	1.000 1	0.432 51	0.828 21	0.673 30	0.788 65	0.552 29

DENet	0.629 43	1.000 1	0.797 37	0.608 46	0.589 32	0.627 56	0.219 32	0.882 1	0.310 50	0.402 53	0.383 56	0.396 46	0.650 23	1.000 1	0.663 26	0.543 64	0.691 28	1.000 1	0.568 26

3D-MPA	0.611 44	1.000 1	0.833 23	0.765 30	0.526 43	0.756 45	0.136 42	0.588 47	0.470 35	0.438 44	0.432 50	0.358 51	0.650 23	0.857 45	0.429 52	0.765 34	0.557 51	1.000 1	0.430 47
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 45	1.000 1	0.801 35	0.599 49	0.535 40	0.728 49	0.286 24	0.436 60	0.679 18	0.491 33	0.433 48	0.256 53	0.404 60	0.857 45	0.620 34	0.724 43	0.510 58	1.000 1	0.539 31

AOIA	0.601 46	1.000 1	0.761 47	0.687 40	0.485 49	0.828 27	0.008 60	0.663 34	0.405 43	0.405 52	0.425 51	0.490 34	0.596 42	0.714 56	0.553 42	0.779 29	0.597 41	0.992 47	0.424 49

PCJC	0.578 47	1.000 1	0.810 29	0.583 53	0.449 53	0.813 33	0.042 54	0.603 45	0.341 49	0.490 34	0.465 45	0.410 44	0.650 23	0.835 53	0.264 62	0.694 51	0.561 49	0.889 58	0.504 37

SSEN	0.575 48	1.000 1	0.761 46	0.473 56	0.477 51	0.795 38	0.066 50	0.529 52	0.658 22	0.460 38	0.461 46	0.380 48	0.331 62	0.859 44	0.401 55	0.692 53	0.653 33	1.000 1	0.348 54
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 49	0.528 66	0.708 57	0.626 44	0.580 34	0.745 48	0.063 51	0.627 41	0.240 54	0.400 54	0.497 42	0.464 36	0.515 49	1.000 1	0.475 47	0.745 39	0.571 46	1.000 1	0.429 48

NeuralBF	0.555 50	0.667 56	0.896 15	0.843 17	0.517 45	0.751 46	0.029 55	0.519 53	0.414 42	0.439 42	0.465 44	0.000 72	0.484 51	0.857 45	0.287 60	0.693 52	0.651 35	1.000 1	0.485 41
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 51	1.000 1	0.807 32	0.588 52	0.327 58	0.647 54	0.004 62	0.815 7	0.180 57	0.418 49	0.364 58	0.182 56	0.445 54	1.000 1	0.442 50	0.688 54	0.571 47	1.000 1	0.396 50
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
ClickSeg_Instance	0.539 52	1.000 1	0.621 61	0.300 59	0.530 42	0.698 50	0.127 43	0.533 51	0.222 55	0.430 47	0.400 53	0.365 49	0.574 46	0.938 41	0.472 48	0.659 56	0.543 54	0.944 53	0.347 55

One_Thing_One_Click	0.529 53	0.667 56	0.718 53	0.777 27	0.399 54	0.683 51	0.000 65	0.669 32	0.138 60	0.391 55	0.374 57	0.539 30	0.360 61	0.641 60	0.556 41	0.774 32	0.593 42	0.997 37	0.251 60
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 54	1.000 1	0.538 66	0.282 60	0.468 52	0.790 39	0.173 34	0.345 62	0.429 39	0.413 51	0.484 43	0.176 57	0.595 44	0.591 61	0.522 44	0.668 55	0.476 59	0.986 51	0.327 56

Occipital-SCS	0.512 55	1.000 1	0.716 54	0.509 55	0.506 47	0.611 57	0.092 47	0.602 46	0.177 58	0.346 58	0.383 55	0.165 58	0.442 55	0.850 52	0.386 57	0.618 60	0.543 55	0.889 58	0.389 51

3D-BoNet	0.488 56	1.000 1	0.672 60	0.590 51	0.301 60	0.484 67	0.098 46	0.620 43	0.306 51	0.341 59	0.259 62	0.125 60	0.434 57	0.796 55	0.402 54	0.499 66	0.513 57	0.909 57	0.439 46
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 57	0.667 56	0.712 56	0.595 50	0.259 63	0.550 63	0.000 65	0.613 44	0.175 59	0.250 64	0.434 47	0.437 40	0.411 59	0.857 45	0.485 46	0.591 63	0.267 69	0.944 53	0.359 53
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 58	0.667 56	0.685 59	0.677 42	0.372 56	0.562 61	0.000 65	0.482 57	0.244 53	0.316 61	0.298 59	0.052 67	0.442 56	0.857 45	0.267 61	0.702 48	0.559 50	1.000 1	0.287 58

SALoss-ResNet	0.459 59	1.000 1	0.737 52	0.159 70	0.259 62	0.587 59	0.138 41	0.475 58	0.217 56	0.416 50	0.408 52	0.128 59	0.315 63	0.714 56	0.411 53	0.536 65	0.590 43	0.873 62	0.304 57
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 60	0.528 66	0.555 64	0.381 57	0.382 55	0.633 55	0.002 63	0.509 55	0.260 52	0.361 57	0.432 49	0.327 52	0.451 53	0.571 62	0.367 58	0.639 58	0.386 60	0.980 52	0.276 59
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
SegGroup_ins	0.445 61	0.667 56	0.773 43	0.185 67	0.317 59	0.656 53	0.000 65	0.407 61	0.134 61	0.381 56	0.267 61	0.217 55	0.476 52	0.714 56	0.452 49	0.629 59	0.514 56	1.000 1	0.222 63
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 62	1.000 1	0.432 69	0.245 62	0.190 64	0.577 60	0.013 59	0.263 64	0.033 67	0.320 60	0.240 63	0.075 63	0.422 58	0.857 45	0.117 67	0.699 49	0.271 68	0.883 60	0.235 62
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
Hier3D	0.323 63	0.667 56	0.542 65	0.264 61	0.157 67	0.550 62	0.000 65	0.205 67	0.009 69	0.270 63	0.218 64	0.075 63	0.500 50	0.688 59	0.007 73	0.698 50	0.301 65	0.459 70	0.200 64
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
UNet-backbone	0.319 64	0.667 56	0.715 55	0.233 63	0.189 65	0.479 68	0.008 60	0.218 65	0.067 66	0.201 66	0.173 65	0.107 61	0.123 68	0.438 64	0.150 64	0.615 61	0.355 61	0.916 56	0.093 72

R-PointNet	0.306 65	0.500 68	0.405 70	0.311 58	0.348 57	0.589 58	0.054 52	0.068 70	0.126 62	0.283 62	0.290 60	0.028 68	0.219 66	0.214 67	0.331 59	0.396 70	0.275 66	0.821 64	0.245 61

Region-18class	0.284 66	0.250 72	0.751 51	0.228 65	0.270 61	0.521 64	0.000 65	0.468 59	0.008 71	0.205 65	0.127 66	0.000 72	0.068 70	0.070 71	0.262 63	0.652 57	0.323 63	0.740 66	0.173 65

SemRegionNet-20cls	0.250 67	0.333 69	0.613 62	0.229 64	0.163 66	0.493 65	0.000 65	0.304 63	0.107 63	0.147 69	0.100 68	0.052 66	0.231 64	0.119 69	0.039 69	0.445 68	0.325 62	0.654 67	0.141 68

3D-BEVIS	0.248 68	0.667 56	0.566 63	0.076 71	0.035 73	0.394 71	0.027 57	0.035 72	0.098 64	0.099 71	0.030 72	0.025 69	0.098 69	0.375 66	0.126 66	0.604 62	0.181 71	0.854 63	0.171 66
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
tmp	0.248 68	0.667 56	0.437 68	0.188 66	0.153 68	0.491 66	0.000 65	0.208 66	0.094 65	0.153 68	0.099 69	0.057 65	0.217 67	0.119 69	0.039 69	0.466 67	0.302 64	0.640 68	0.140 69

Sem_Recon_ins	0.227 70	0.764 55	0.486 67	0.069 72	0.098 70	0.426 70	0.017 58	0.067 71	0.015 68	0.172 67	0.100 67	0.096 62	0.054 72	0.183 68	0.135 65	0.366 71	0.260 70	0.614 69	0.168 67

ASIS	0.199 71	0.333 69	0.253 72	0.167 69	0.140 69	0.438 69	0.000 65	0.177 68	0.008 70	0.121 70	0.069 70	0.004 71	0.231 65	0.429 65	0.036 71	0.445 69	0.273 67	0.333 72	0.119 71

Sgpn_scannet	0.143 72	0.208 73	0.390 71	0.169 68	0.065 71	0.275 72	0.029 56	0.069 69	0.000 72	0.087 72	0.043 71	0.014 70	0.027 73	0.000 72	0.112 68	0.351 72	0.168 72	0.438 71	0.138 70

MaskRCNN 2d->3d Proj	0.058 73	0.333 69	0.002 73	0.000 73	0.053 72	0.002 73	0.002 64	0.021 73	0.000 72	0.045 73	0.024 73	0.238 54	0.065 71	0.000 72	0.014 72	0.107 73	0.020 73	0.110 73	0.006 73

3D Semantic Instance Benchmark