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

Spherical Mask(CtoF)	0.812 1	1.000 1	0.973 3	0.852 12	0.718 4	0.917 5	0.574 4	0.677 25	0.748 8	0.729 8	0.715 5	0.795 2	0.809 1	1.000 1	0.831 2	0.854 8	0.787 7	1.000 1	0.638 4

SIM3D	0.805 2	1.000 1	0.971 4	0.863 11	0.686 13	0.924 4	0.552 7	0.739 17	0.674 15	0.740 6	0.666 11	0.807 1	0.789 7	1.000 1	0.803 5	0.866 5	0.775 13	1.000 1	0.639 3

OneFormer3D	0.801 3	1.000 1	0.973 2	0.909 5	0.698 10	0.928 2	0.582 3	0.668 29	0.685 13	0.780 2	0.687 9	0.698 13	0.702 12	1.000 1	0.794 7	0.900 2	0.784 9	0.986 46	0.635 5
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
UniPerception	0.800 4	1.000 1	0.930 6	0.872 9	0.727 3	0.862 18	0.454 13	0.764 13	0.820 1	0.746 5	0.706 7	0.750 3	0.772 8	0.926 39	0.764 12	0.818 23	0.826 1	0.997 34	0.660 2

InsSSM	0.799 5	1.000 1	0.915 8	0.710 35	0.729 2	0.925 3	0.664 1	0.670 27	0.770 5	0.766 3	0.739 2	0.737 4	0.700 13	1.000 1	0.792 8	0.829 17	0.815 3	0.997 34	0.625 7

TST3D	0.795 6	1.000 1	0.929 7	0.918 4	0.709 7	0.884 13	0.596 2	0.704 22	0.769 6	0.734 7	0.644 15	0.699 12	0.751 10	1.000 1	0.794 6	0.876 4	0.757 17	0.997 34	0.550 26

ExtMask3D	0.789 7	1.000 1	0.988 1	0.756 28	0.706 8	0.912 6	0.429 14	0.647 34	0.806 4	0.755 4	0.673 10	0.689 14	0.772 9	1.000 1	0.789 9	0.852 9	0.811 4	1.000 1	0.617 10

Queryformer	0.787 8	1.000 1	0.933 5	0.601 44	0.754 1	0.886 11	0.558 6	0.661 31	0.767 7	0.665 13	0.716 4	0.639 19	0.808 3	1.000 1	0.844 1	0.897 3	0.804 5	1.000 1	0.624 8

MAFT	0.786 9	1.000 1	0.894 13	0.807 18	0.694 12	0.893 9	0.486 9	0.674 26	0.740 9	0.786 1	0.704 8	0.727 6	0.739 11	1.000 1	0.707 18	0.849 11	0.756 18	1.000 1	0.685 1

Mask3D	0.780 10	1.000 1	0.786 37	0.716 33	0.696 11	0.885 12	0.500 8	0.714 20	0.810 3	0.672 12	0.715 5	0.679 15	0.809 1	1.000 1	0.831 2	0.833 15	0.787 7	1.000 1	0.602 14
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
SPFormer	0.770 11	0.903 50	0.903 10	0.806 19	0.609 26	0.886 10	0.568 5	0.815 6	0.705 12	0.711 9	0.655 12	0.652 18	0.685 18	1.000 1	0.789 10	0.809 24	0.776 12	1.000 1	0.583 19
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
SoftGroup++	0.769 12	1.000 1	0.803 30	0.937 1	0.684 14	0.865 15	0.213 29	0.870 2	0.664 17	0.571 19	0.758 1	0.702 10	0.807 4	1.000 1	0.653 25	0.902 1	0.792 6	1.000 1	0.626 6

SoftGroup	0.761 13	1.000 1	0.808 26	0.845 13	0.716 5	0.862 17	0.243 26	0.824 4	0.655 19	0.620 14	0.734 3	0.699 11	0.791 6	0.981 33	0.716 16	0.844 12	0.769 14	1.000 1	0.594 17
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 14	1.000 1	0.904 9	0.731 31	0.678 15	0.895 7	0.458 11	0.644 36	0.670 16	0.710 10	0.620 20	0.732 5	0.650 20	1.000 1	0.756 13	0.778 27	0.779 10	1.000 1	0.614 11
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 15	1.000 1	0.774 38	0.867 10	0.621 22	0.934 1	0.404 15	0.706 21	0.812 2	0.605 17	0.633 18	0.626 20	0.690 17	1.000 1	0.640 27	0.820 20	0.777 11	1.000 1	0.612 12
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
PBNet	0.747 16	1.000 1	0.818 22	0.837 15	0.713 6	0.844 20	0.457 12	0.647 34	0.711 11	0.614 15	0.617 22	0.657 17	0.650 20	1.000 1	0.692 19	0.822 19	0.765 16	1.000 1	0.595 16
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 17	1.000 1	0.788 35	0.724 32	0.642 20	0.859 19	0.248 25	0.787 11	0.618 22	0.596 18	0.653 14	0.722 8	0.583 41	1.000 1	0.766 11	0.861 6	0.825 2	1.000 1	0.504 32

IPCA-Inst	0.731 18	1.000 1	0.788 36	0.884 8	0.698 9	0.788 36	0.252 24	0.760 14	0.646 20	0.511 27	0.637 17	0.665 16	0.804 5	1.000 1	0.644 26	0.778 28	0.747 20	1.000 1	0.561 23

TopoSeg	0.725 19	1.000 1	0.806 29	0.933 2	0.668 17	0.758 40	0.272 23	0.734 19	0.630 21	0.549 23	0.654 13	0.606 21	0.697 16	0.966 36	0.612 31	0.839 13	0.754 19	1.000 1	0.573 20

DKNet	0.718 20	1.000 1	0.814 23	0.782 22	0.619 23	0.872 14	0.224 27	0.751 16	0.569 26	0.677 11	0.585 26	0.724 7	0.633 31	0.981 33	0.515 41	0.819 21	0.736 21	1.000 1	0.617 9
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
SSEC	0.707 21	1.000 1	0.850 15	0.924 3	0.648 18	0.747 43	0.162 31	0.862 3	0.572 25	0.520 25	0.624 19	0.549 24	0.649 29	1.000 1	0.560 36	0.706 43	0.768 15	1.000 1	0.591 18

HAIS	0.699 22	1.000 1	0.849 16	0.820 16	0.675 16	0.808 30	0.279 21	0.757 15	0.465 32	0.517 26	0.596 24	0.559 23	0.600 35	1.000 1	0.654 24	0.767 30	0.676 25	0.994 42	0.560 24
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
SSTNet	0.698 23	1.000 1	0.697 54	0.888 7	0.556 33	0.803 31	0.387 16	0.626 38	0.417 37	0.556 22	0.585 27	0.702 9	0.600 35	1.000 1	0.824 4	0.720 42	0.692 23	1.000 1	0.509 31
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
DualGroup	0.694 24	1.000 1	0.799 32	0.811 17	0.622 21	0.817 25	0.376 17	0.805 9	0.590 24	0.487 31	0.568 30	0.525 28	0.650 20	0.835 49	0.600 32	0.829 16	0.655 28	1.000 1	0.526 28

SphereSeg	0.680 25	1.000 1	0.856 14	0.744 29	0.618 24	0.893 8	0.151 32	0.651 33	0.713 10	0.537 24	0.579 29	0.430 38	0.651 19	1.000 1	0.389 52	0.744 37	0.697 22	0.991 44	0.601 15

DANCENET	0.680 25	1.000 1	0.807 27	0.733 30	0.600 27	0.768 39	0.375 18	0.543 46	0.538 27	0.610 16	0.599 23	0.498 29	0.632 33	0.981 33	0.739 15	0.856 7	0.633 34	0.882 57	0.454 41

Box2Mask	0.677 27	1.000 1	0.847 17	0.771 24	0.509 42	0.816 26	0.277 22	0.558 45	0.482 29	0.562 21	0.640 16	0.448 34	0.700 13	1.000 1	0.666 20	0.852 10	0.578 41	0.997 34	0.488 36
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 28	1.000 1	0.758 46	0.682 37	0.576 31	0.842 21	0.477 10	0.504 52	0.524 28	0.567 20	0.585 28	0.451 33	0.557 43	1.000 1	0.751 14	0.797 25	0.563 44	1.000 1	0.467 40
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
Mask-Group	0.664 29	1.000 1	0.822 21	0.764 27	0.616 25	0.815 27	0.139 36	0.694 24	0.597 23	0.459 35	0.566 31	0.599 22	0.600 35	0.516 59	0.715 17	0.819 22	0.635 32	1.000 1	0.603 13
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 30	1.000 1	0.760 44	0.667 39	0.581 29	0.863 16	0.323 19	0.655 32	0.477 30	0.473 33	0.549 33	0.432 37	0.650 20	1.000 1	0.655 23	0.738 38	0.585 40	0.944 49	0.472 39

CSC-Pretrained	0.648 31	1.000 1	0.810 24	0.768 25	0.523 40	0.813 28	0.143 35	0.819 5	0.389 40	0.422 44	0.511 37	0.443 35	0.650 20	1.000 1	0.624 29	0.732 39	0.634 33	1.000 1	0.375 48

PE	0.645 32	1.000 1	0.773 40	0.798 21	0.538 35	0.786 37	0.088 44	0.799 10	0.350 44	0.435 42	0.547 34	0.545 25	0.646 30	0.933 38	0.562 35	0.761 33	0.556 49	0.997 34	0.501 34
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
RPGN	0.643 33	1.000 1	0.758 45	0.582 50	0.539 34	0.826 24	0.046 49	0.765 12	0.372 42	0.436 41	0.588 25	0.539 27	0.650 20	1.000 1	0.577 33	0.750 35	0.653 30	0.997 34	0.495 35
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
Dyco3D	0.641 34	1.000 1	0.841 18	0.893 6	0.531 37	0.802 32	0.115 41	0.588 43	0.448 34	0.438 39	0.537 36	0.430 39	0.550 44	0.857 41	0.534 39	0.764 32	0.657 27	0.987 45	0.568 21
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
GICN	0.638 35	1.000 1	0.895 12	0.800 20	0.480 46	0.676 48	0.144 34	0.737 18	0.354 43	0.447 36	0.400 50	0.365 45	0.700 13	1.000 1	0.569 34	0.836 14	0.599 36	1.000 1	0.473 38

PointGroup	0.636 36	1.000 1	0.765 41	0.624 41	0.505 44	0.797 33	0.116 40	0.696 23	0.384 41	0.441 37	0.559 32	0.476 31	0.596 38	1.000 1	0.666 20	0.756 34	0.556 48	0.997 34	0.513 30
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 37	0.667 52	0.797 34	0.714 34	0.562 32	0.774 38	0.146 33	0.810 8	0.429 36	0.476 32	0.546 35	0.399 41	0.633 31	1.000 1	0.632 28	0.722 41	0.609 35	1.000 1	0.514 29
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 38	1.000 1	0.829 20	0.606 43	0.646 19	0.836 22	0.068 45	0.511 50	0.462 33	0.507 28	0.619 21	0.389 43	0.610 34	1.000 1	0.432 47	0.828 18	0.673 26	0.788 61	0.552 25

DENet	0.629 39	1.000 1	0.797 33	0.608 42	0.589 28	0.627 52	0.219 28	0.882 1	0.310 46	0.402 49	0.383 52	0.396 42	0.650 20	1.000 1	0.663 22	0.543 60	0.691 24	1.000 1	0.568 22

3D-MPA	0.611 40	1.000 1	0.833 19	0.765 26	0.526 39	0.756 41	0.136 38	0.588 43	0.470 31	0.438 40	0.432 46	0.358 47	0.650 20	0.857 41	0.429 48	0.765 31	0.557 47	1.000 1	0.430 43
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 41	1.000 1	0.801 31	0.599 45	0.535 36	0.728 45	0.286 20	0.436 56	0.679 14	0.491 29	0.433 44	0.256 49	0.404 56	0.857 41	0.620 30	0.724 40	0.510 54	1.000 1	0.539 27

AOIA	0.601 42	1.000 1	0.761 43	0.687 36	0.485 45	0.828 23	0.008 56	0.663 30	0.405 39	0.405 48	0.425 47	0.490 30	0.596 38	0.714 52	0.553 38	0.779 26	0.597 37	0.992 43	0.424 45

PCJC	0.578 43	1.000 1	0.810 25	0.583 49	0.449 49	0.813 29	0.042 50	0.603 41	0.341 45	0.490 30	0.465 41	0.410 40	0.650 20	0.835 49	0.264 58	0.694 47	0.561 45	0.889 54	0.504 33

SSEN	0.575 44	1.000 1	0.761 42	0.473 52	0.477 47	0.795 34	0.066 46	0.529 48	0.658 18	0.460 34	0.461 42	0.380 44	0.331 58	0.859 40	0.401 51	0.692 49	0.653 29	1.000 1	0.348 50
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 45	0.528 62	0.708 53	0.626 40	0.580 30	0.745 44	0.063 47	0.627 37	0.240 50	0.400 50	0.497 38	0.464 32	0.515 45	1.000 1	0.475 43	0.745 36	0.571 42	1.000 1	0.429 44

NeuralBF	0.555 46	0.667 52	0.896 11	0.843 14	0.517 41	0.751 42	0.029 51	0.519 49	0.414 38	0.439 38	0.465 40	0.000 68	0.484 47	0.857 41	0.287 56	0.693 48	0.651 31	1.000 1	0.485 37
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 47	1.000 1	0.807 28	0.588 48	0.327 54	0.647 50	0.004 58	0.815 7	0.180 53	0.418 45	0.364 54	0.182 52	0.445 50	1.000 1	0.442 46	0.688 50	0.571 43	1.000 1	0.396 46
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
ClickSeg_Instance	0.539 48	1.000 1	0.621 57	0.300 55	0.530 38	0.698 46	0.127 39	0.533 47	0.222 51	0.430 43	0.400 49	0.365 45	0.574 42	0.938 37	0.472 44	0.659 52	0.543 50	0.944 49	0.347 51

One_Thing_One_Click	0.529 49	0.667 52	0.718 49	0.777 23	0.399 50	0.683 47	0.000 61	0.669 28	0.138 56	0.391 51	0.374 53	0.539 26	0.360 57	0.641 56	0.556 37	0.774 29	0.593 38	0.997 34	0.251 56
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 50	1.000 1	0.538 62	0.282 56	0.468 48	0.790 35	0.173 30	0.345 58	0.429 35	0.413 47	0.484 39	0.176 53	0.595 40	0.591 57	0.522 40	0.668 51	0.476 55	0.986 47	0.327 52

Occipital-SCS	0.512 51	1.000 1	0.716 50	0.509 51	0.506 43	0.611 53	0.092 43	0.602 42	0.177 54	0.346 54	0.383 51	0.165 54	0.442 51	0.850 48	0.386 53	0.618 56	0.543 51	0.889 54	0.389 47

3D-BoNet	0.488 52	1.000 1	0.672 56	0.590 47	0.301 56	0.484 63	0.098 42	0.620 39	0.306 47	0.341 55	0.259 58	0.125 56	0.434 53	0.796 51	0.402 50	0.499 62	0.513 53	0.909 53	0.439 42
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 53	0.667 52	0.712 52	0.595 46	0.259 59	0.550 59	0.000 61	0.613 40	0.175 55	0.250 60	0.434 43	0.437 36	0.411 55	0.857 41	0.485 42	0.591 59	0.267 65	0.944 49	0.359 49
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 54	0.667 52	0.685 55	0.677 38	0.372 52	0.562 57	0.000 61	0.482 53	0.244 49	0.316 57	0.298 55	0.052 63	0.442 52	0.857 41	0.267 57	0.702 44	0.559 46	1.000 1	0.287 54

SALoss-ResNet	0.459 55	1.000 1	0.737 48	0.159 66	0.259 58	0.587 55	0.138 37	0.475 54	0.217 52	0.416 46	0.408 48	0.128 55	0.315 59	0.714 52	0.411 49	0.536 61	0.590 39	0.873 58	0.304 53
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 56	0.528 62	0.555 60	0.381 53	0.382 51	0.633 51	0.002 59	0.509 51	0.260 48	0.361 53	0.432 45	0.327 48	0.451 49	0.571 58	0.367 54	0.639 54	0.386 56	0.980 48	0.276 55
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
SegGroup_ins	0.445 57	0.667 52	0.773 39	0.185 63	0.317 55	0.656 49	0.000 61	0.407 57	0.134 57	0.381 52	0.267 57	0.217 51	0.476 48	0.714 52	0.452 45	0.629 55	0.514 52	1.000 1	0.222 59
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 58	1.000 1	0.432 65	0.245 58	0.190 60	0.577 56	0.013 55	0.263 60	0.033 63	0.320 56	0.240 59	0.075 59	0.422 54	0.857 41	0.117 63	0.699 45	0.271 64	0.883 56	0.235 58
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
Hier3D	0.323 59	0.667 52	0.542 61	0.264 57	0.157 63	0.550 58	0.000 61	0.205 63	0.009 65	0.270 59	0.218 60	0.075 59	0.500 46	0.688 55	0.007 69	0.698 46	0.301 61	0.459 66	0.200 60
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
UNet-backbone	0.319 60	0.667 52	0.715 51	0.233 59	0.189 61	0.479 64	0.008 56	0.218 61	0.067 62	0.201 62	0.173 61	0.107 57	0.123 64	0.438 60	0.150 60	0.615 57	0.355 57	0.916 52	0.093 68

R-PointNet	0.306 61	0.500 64	0.405 66	0.311 54	0.348 53	0.589 54	0.054 48	0.068 66	0.126 58	0.283 58	0.290 56	0.028 64	0.219 62	0.214 63	0.331 55	0.396 66	0.275 62	0.821 60	0.245 57

Region-18class	0.284 62	0.250 68	0.751 47	0.228 61	0.270 57	0.521 60	0.000 61	0.468 55	0.008 67	0.205 61	0.127 62	0.000 68	0.068 66	0.070 67	0.262 59	0.652 53	0.323 59	0.740 62	0.173 61

SemRegionNet-20cls	0.250 63	0.333 65	0.613 58	0.229 60	0.163 62	0.493 61	0.000 61	0.304 59	0.107 59	0.147 65	0.100 64	0.052 62	0.231 60	0.119 65	0.039 65	0.445 64	0.325 58	0.654 63	0.141 64

tmp	0.248 64	0.667 52	0.437 64	0.188 62	0.153 64	0.491 62	0.000 61	0.208 62	0.094 61	0.153 64	0.099 65	0.057 61	0.217 63	0.119 65	0.039 65	0.466 63	0.302 60	0.640 64	0.140 65

3D-BEVIS	0.248 64	0.667 52	0.566 59	0.076 67	0.035 69	0.394 67	0.027 53	0.035 68	0.098 60	0.099 67	0.030 68	0.025 65	0.098 65	0.375 62	0.126 62	0.604 58	0.181 67	0.854 59	0.171 62
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
Sem_Recon_ins	0.227 66	0.764 51	0.486 63	0.069 68	0.098 66	0.426 66	0.017 54	0.067 67	0.015 64	0.172 63	0.100 63	0.096 58	0.054 68	0.183 64	0.135 61	0.366 67	0.260 66	0.614 65	0.168 63

ASIS	0.199 67	0.333 65	0.253 68	0.167 65	0.140 65	0.438 65	0.000 61	0.177 64	0.008 66	0.121 66	0.069 66	0.004 67	0.231 61	0.429 61	0.036 67	0.445 65	0.273 63	0.333 68	0.119 67

Sgpn_scannet	0.143 68	0.208 69	0.390 67	0.169 64	0.065 67	0.275 68	0.029 52	0.069 65	0.000 68	0.087 68	0.043 67	0.014 66	0.027 69	0.000 68	0.112 64	0.351 68	0.168 68	0.438 67	0.138 66

MaskRCNN 2d->3d Proj	0.058 69	0.333 65	0.002 69	0.000 69	0.053 68	0.002 69	0.002 60	0.021 69	0.000 68	0.045 69	0.024 69	0.238 50	0.065 67	0.000 68	0.014 68	0.107 69	0.020 69	0.110 69	0.006 69

3D Semantic Instance Benchmark