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 25%	bathtub	bed	bookshelf	cabinet	chair	counter	curtain	desk	door	otherfurniture	picture	refrigerator	shower curtain	sink	sofa	table	toilet	window

DD-UNet+Group	0.764 36	1.000 1	0.897 47	0.837 21	0.753 32	0.830 40	0.459 44	0.824 16	0.699 45	0.629 36	0.653 27	0.438 48	0.650 23	1.000 1	0.880 28	0.858 32	0.690 47	1.000 1	0.650 34
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
DKNet	0.815 22	1.000 1	0.930 32	0.844 20	0.765 28	0.915 14	0.534 37	0.805 22	0.805 29	0.807 14	0.654 26	0.763 9	0.650 23	1.000 1	0.794 45	0.881 15	0.766 24	1.000 1	0.758 18
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
SSEC	0.820 21	1.000 1	0.983 12	0.924 3	0.826 10	0.817 44	0.415 48	0.899 4	0.793 31	0.673 30	0.731 13	0.636 25	0.653 21	1.000 1	0.939 7	0.804 52	0.878 7	1.000 1	0.780 13

SegGroup_ins	0.637 55	1.000 1	0.923 39	0.593 60	0.561 58	0.746 53	0.143 65	0.504 57	0.766 37	0.485 59	0.442 57	0.372 51	0.530 50	0.714 55	0.815 41	0.775 57	0.673 49	1.000 1	0.431 58
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
OccuSeg+instance	0.742 39	1.000 1	0.923 37	0.785 35	0.745 36	0.867 26	0.557 31	0.578 52	0.729 41	0.670 31	0.644 31	0.488 43	0.577 46	1.000 1	0.794 45	0.830 43	0.620 55	1.000 1	0.550 48
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
PBNet	0.825 20	1.000 1	0.963 22	0.837 22	0.843 6	0.865 27	0.822 3	0.647 43	0.878 16	0.733 19	0.639 32	0.683 20	0.650 23	1.000 1	0.853 33	0.870 25	0.820 20	1.000 1	0.744 21
W.Zhao, Y.Yan, C.Yang, J.Ye,X.Yang,K.Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
RWSeg	0.739 40	1.000 1	0.899 46	0.759 44	0.753 33	0.823 42	0.282 54	0.691 39	0.658 48	0.582 47	0.594 40	0.547 34	0.628 38	1.000 1	0.795 44	0.868 27	0.728 36	1.000 1	0.692 29

SoftGroup	0.865 10	1.000 1	0.969 19	0.860 17	0.860 3	0.913 15	0.558 30	0.899 3	0.911 9	0.760 16	0.828 1	0.736 14	0.802 6	0.981 37	0.919 13	0.875 20	0.877 10	1.000 1	0.820 7
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral]
IPCA-Inst	0.851 13	1.000 1	0.968 20	0.884 11	0.842 7	0.862 32	0.693 18	0.812 20	0.888 14	0.677 29	0.783 3	0.698 18	0.807 5	1.000 1	0.911 19	0.865 29	0.865 13	1.000 1	0.757 19

SoftGroup++	0.874 6	1.000 1	0.972 18	0.947 1	0.839 8	0.898 19	0.556 33	0.913 2	0.881 15	0.756 17	0.828 2	0.748 12	0.821 1	1.000 1	0.937 9	0.937 1	0.887 3	1.000 1	0.821 6

Spherical Mask(CtoF)	0.875 4	1.000 1	0.991 9	0.873 14	0.850 5	0.946 5	0.691 19	0.752 29	0.926 4	0.889 5	0.759 7	0.794 4	0.820 2	1.000 1	0.912 14	0.900 8	0.878 8	1.000 1	0.769 15

DENet	0.786 30	1.000 1	0.929 33	0.736 47	0.750 35	0.720 57	0.755 12	0.934 1	0.794 30	0.590 44	0.561 45	0.537 38	0.650 23	1.000 1	0.882 25	0.804 53	0.789 21	1.000 1	0.719 25

MTML	0.731 42	1.000 1	0.992 7	0.779 41	0.609 54	0.746 52	0.308 53	0.867 8	0.601 54	0.607 40	0.539 49	0.519 41	0.550 47	1.000 1	0.824 38	0.869 26	0.729 35	1.000 1	0.616 40
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
CSC-Pretrained	0.791 27	1.000 1	0.996 3	0.829 25	0.767 26	0.889 23	0.600 25	0.819 18	0.770 36	0.594 43	0.620 36	0.541 37	0.700 15	1.000 1	0.941 5	0.889 12	0.763 25	1.000 1	0.526 52

SSTNet	0.789 28	1.000 1	0.840 55	0.888 10	0.717 42	0.835 37	0.717 15	0.684 41	0.627 51	0.724 20	0.652 28	0.727 16	0.600 40	1.000 1	0.912 14	0.822 45	0.757 28	1.000 1	0.691 30
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
Mask-Group	0.792 26	1.000 1	0.968 21	0.812 27	0.766 27	0.864 28	0.460 42	0.815 19	0.888 13	0.598 42	0.651 29	0.639 23	0.600 40	0.918 43	0.941 5	0.896 10	0.721 37	1.000 1	0.723 24
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022
GICN	0.788 29	1.000 1	0.978 15	0.867 16	0.781 23	0.833 38	0.527 38	0.824 16	0.806 28	0.549 51	0.596 39	0.551 33	0.700 15	1.000 1	0.853 33	0.935 2	0.733 34	1.000 1	0.651 33

SPFormer	0.851 13	1.000 1	0.994 5	0.806 32	0.774 24	0.942 6	0.637 22	0.849 12	0.859 19	0.889 4	0.720 15	0.730 15	0.665 20	1.000 1	0.911 19	0.868 28	0.873 12	1.000 1	0.796 11
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
SPG_WSIS	0.678 53	1.000 1	0.880 51	0.836 23	0.701 46	0.727 56	0.273 58	0.607 47	0.706 44	0.541 54	0.515 53	0.174 59	0.600 40	0.857 46	0.716 52	0.846 39	0.711 39	1.000 1	0.506 53

TopoSeg	0.832 18	1.000 1	0.981 13	0.933 2	0.819 12	0.826 41	0.524 39	0.841 13	0.811 27	0.681 28	0.759 8	0.687 19	0.727 12	0.981 37	0.911 19	0.883 14	0.853 15	1.000 1	0.756 20

ISBNet	0.835 16	1.000 1	0.950 26	0.731 48	0.819 11	0.918 13	0.790 8	0.740 31	0.851 21	0.831 10	0.661 25	0.742 13	0.650 23	1.000 1	0.937 8	0.814 50	0.836 17	1.000 1	0.765 17
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
ExtMask3D	0.867 9	1.000 1	1.000 1	0.756 46	0.816 14	0.940 7	0.795 7	0.760 28	0.862 17	0.888 6	0.739 11	0.763 8	0.774 8	1.000 1	0.929 12	0.878 18	0.879 6	1.000 1	0.819 8

3D-MPA	0.737 41	1.000 1	0.933 31	0.785 35	0.794 20	0.831 39	0.279 56	0.588 49	0.695 46	0.616 37	0.559 46	0.556 32	0.650 23	1.000 1	0.809 42	0.875 21	0.696 43	1.000 1	0.608 44
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020
Queryformer	0.874 6	1.000 1	0.978 16	0.809 31	0.876 1	0.936 10	0.702 16	0.716 34	0.920 7	0.875 8	0.766 5	0.772 6	0.818 4	1.000 1	0.995 2	0.916 5	0.892 2	1.000 1	0.767 16

MAFT	0.860 11	1.000 1	0.990 10	0.810 30	0.829 9	0.949 4	0.809 5	0.688 40	0.836 22	0.904 2	0.751 10	0.796 3	0.741 11	1.000 1	0.864 32	0.848 37	0.837 16	1.000 1	0.828 3

OSIS	0.725 43	1.000 1	0.885 50	0.653 56	0.657 51	0.801 45	0.576 29	0.695 38	0.828 25	0.698 25	0.534 50	0.457 47	0.500 54	0.857 46	0.831 37	0.841 40	0.627 53	1.000 1	0.619 39

SIM3D	0.852 12	1.000 1	0.842 54	0.913 5	0.804 16	0.922 12	0.828 2	0.781 25	0.909 10	0.831 11	0.690 23	0.755 10	0.676 19	1.000 1	0.997 1	0.848 38	0.735 33	1.000 1	0.815 9

TD3D	0.875 4	1.000 1	0.976 17	0.877 12	0.783 22	0.970 1	0.889 1	0.828 15	0.945 3	0.803 15	0.713 16	0.720 17	0.709 13	1.000 1	0.936 10	0.934 3	0.873 11	1.000 1	0.791 12
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
GraphCut	0.832 18	1.000 1	0.922 40	0.724 50	0.798 17	0.902 18	0.701 17	0.856 10	0.859 18	0.715 22	0.706 17	0.748 11	0.640 34	1.000 1	0.934 11	0.862 31	0.880 5	1.000 1	0.729 22

SSEN	0.724 44	1.000 1	0.926 35	0.781 40	0.661 49	0.845 36	0.596 28	0.529 55	0.764 39	0.653 33	0.489 56	0.461 46	0.500 54	0.859 45	0.765 48	0.872 23	0.761 27	1.000 1	0.577 46
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv
DualGroup	0.782 32	1.000 1	0.927 34	0.811 28	0.772 25	0.853 34	0.631 24	0.805 22	0.773 33	0.613 39	0.611 37	0.610 28	0.650 23	0.835 54	0.881 26	0.879 17	0.750 31	1.000 1	0.675 31

NeuralBF	0.718 45	1.000 1	0.945 27	0.901 7	0.754 31	0.817 43	0.460 42	0.700 37	0.772 34	0.688 26	0.568 44	0.000 67	0.500 54	0.981 37	0.606 58	0.872 22	0.740 32	1.000 1	0.614 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
TST3D	0.879 3	1.000 1	0.994 4	0.921 4	0.807 15	0.939 8	0.771 10	0.887 6	0.923 6	0.862 9	0.722 14	0.768 7	0.756 9	1.000 1	0.910 22	0.904 6	0.836 18	0.999 33	0.824 5

UniPerception	0.884 2	1.000 1	0.979 14	0.872 15	0.869 2	0.892 20	0.806 6	0.890 5	0.835 23	0.892 3	0.755 9	0.811 1	0.779 7	0.955 40	0.951 4	0.876 19	0.914 1	0.997 34	0.840 2

PointGroup	0.778 33	1.000 1	0.900 44	0.798 33	0.715 43	0.863 29	0.493 40	0.706 36	0.895 11	0.569 49	0.701 18	0.576 31	0.639 35	1.000 1	0.880 28	0.851 35	0.719 38	0.997 34	0.709 27
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]
PE	0.776 34	1.000 1	0.900 45	0.860 17	0.728 40	0.869 25	0.400 49	0.857 9	0.774 32	0.568 50	0.701 19	0.602 29	0.646 32	0.933 42	0.843 36	0.890 11	0.691 46	0.997 34	0.709 26
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
Box2Mask	0.803 24	1.000 1	0.962 24	0.874 13	0.707 45	0.887 24	0.686 20	0.598 48	0.961 1	0.715 23	0.694 21	0.469 45	0.700 15	1.000 1	0.912 14	0.902 7	0.753 29	0.997 34	0.637 36
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022
RPGN	0.806 23	1.000 1	0.992 7	0.789 34	0.723 41	0.891 21	0.650 21	0.810 21	0.832 24	0.665 32	0.699 20	0.658 21	0.700 15	1.000 1	0.881 26	0.832 42	0.774 22	0.997 34	0.613 42
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
One_Thing_One_Click	0.675 54	1.000 1	0.823 56	0.782 38	0.621 52	0.766 49	0.211 60	0.736 33	0.560 58	0.586 45	0.522 51	0.636 26	0.453 58	0.641 58	0.853 33	0.850 36	0.694 44	0.997 34	0.411 59
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
OneFormer3D	0.896 1	1.000 1	1.000 1	0.913 5	0.858 4	0.951 3	0.786 9	0.837 14	0.916 8	0.908 1	0.778 4	0.803 2	0.750 10	1.000 1	0.976 3	0.926 4	0.882 4	0.995 40	0.849 1
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
HAIS	0.803 24	1.000 1	0.994 5	0.820 26	0.759 29	0.855 33	0.554 34	0.882 7	0.827 26	0.615 38	0.676 24	0.638 24	0.646 32	1.000 1	0.912 14	0.797 55	0.767 23	0.994 41	0.726 23
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
AOIA	0.767 35	1.000 1	0.937 28	0.810 29	0.740 37	0.906 16	0.550 35	0.800 24	0.706 43	0.577 48	0.624 34	0.544 36	0.596 45	0.857 46	0.879 30	0.880 16	0.750 30	0.992 42	0.658 32

SphereSeg	0.835 16	1.000 1	0.963 23	0.891 9	0.794 19	0.954 2	0.822 4	0.710 35	0.961 2	0.721 21	0.693 22	0.530 40	0.653 22	1.000 1	0.867 31	0.857 33	0.859 14	0.991 43	0.771 14

Dyco3D	0.761 38	1.000 1	0.935 30	0.893 8	0.752 34	0.863 30	0.600 25	0.588 49	0.742 40	0.641 34	0.633 33	0.546 35	0.550 47	0.857 46	0.789 47	0.853 34	0.762 26	0.987 44	0.699 28
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
Sparse R-CNN	0.714 46	1.000 1	0.926 36	0.694 51	0.699 47	0.890 22	0.636 23	0.516 56	0.693 47	0.743 18	0.588 41	0.369 52	0.601 39	0.594 60	0.800 43	0.886 13	0.676 48	0.986 45	0.546 49

Mask3D	0.870 8	1.000 1	0.985 11	0.782 39	0.818 13	0.938 9	0.760 11	0.749 30	0.923 5	0.877 7	0.760 6	0.785 5	0.820 2	1.000 1	0.912 14	0.864 30	0.878 8	0.983 46	0.825 4
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
MASC	0.615 56	0.711 63	0.802 58	0.540 62	0.757 30	0.777 48	0.029 66	0.577 53	0.588 57	0.521 57	0.600 38	0.436 49	0.534 49	0.697 56	0.616 57	0.838 41	0.526 57	0.980 47	0.534 51
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
UNet-backbone	0.605 57	1.000 1	0.909 42	0.764 43	0.603 55	0.704 58	0.415 47	0.301 62	0.548 59	0.461 60	0.394 58	0.267 54	0.386 60	0.857 46	0.649 56	0.817 47	0.504 59	0.959 48	0.356 62

R-PointNet	0.544 59	0.500 66	0.655 65	0.661 54	0.663 48	0.765 50	0.432 46	0.214 65	0.612 52	0.584 46	0.499 55	0.204 58	0.286 64	0.429 63	0.655 55	0.650 66	0.539 56	0.950 49	0.499 55

ClickSeg_Instance	0.685 51	1.000 1	0.818 57	0.600 59	0.715 44	0.795 46	0.557 31	0.533 54	0.591 56	0.601 41	0.519 52	0.429 50	0.638 36	0.938 41	0.706 53	0.817 48	0.624 54	0.944 50	0.502 54

PanopticFusion-inst	0.693 48	1.000 1	0.852 53	0.655 55	0.616 53	0.788 47	0.334 51	0.763 27	0.771 35	0.457 61	0.555 47	0.652 22	0.518 51	0.857 46	0.765 48	0.732 61	0.631 51	0.944 50	0.577 47
Gaku Narita, Takashi Seno, Tomoya Ishikawa, Yohsuke Kaji: PanopticFusion: Online Volumetric Semantic Mapping at the Level of Stuff and Things. IROS 2019 (to appear)
DANCENET	0.786 30	1.000 1	0.936 29	0.783 37	0.737 38	0.852 35	0.742 14	0.647 43	0.765 38	0.811 13	0.624 35	0.579 30	0.632 37	1.000 1	0.909 23	0.898 9	0.696 42	0.944 50	0.601 45

INS-Conv-instance	0.762 37	1.000 1	0.923 37	0.765 42	0.785 21	0.905 17	0.600 25	0.655 42	0.646 50	0.683 27	0.647 30	0.530 39	0.650 23	1.000 1	0.824 38	0.830 43	0.693 45	0.944 50	0.644 35

Mask3D_evaluation	0.843 15	1.000 1	0.955 25	0.847 19	0.795 18	0.932 11	0.750 13	0.780 26	0.891 12	0.818 12	0.737 12	0.633 27	0.703 14	1.000 1	0.902 24	0.870 24	0.820 19	0.941 54	0.805 10

Hier3D	0.540 60	1.000 1	0.727 60	0.626 57	0.467 64	0.693 59	0.200 61	0.412 58	0.480 63	0.528 55	0.318 62	0.077 66	0.600 40	0.688 57	0.382 61	0.768 59	0.472 61	0.941 54	0.350 63
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
3D-SIS	0.558 58	1.000 1	0.773 59	0.614 58	0.503 61	0.691 60	0.200 61	0.412 58	0.498 62	0.546 53	0.311 63	0.103 63	0.600 40	0.857 46	0.382 61	0.799 54	0.445 65	0.938 56	0.371 60
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
SALoss-ResNet	0.695 47	1.000 1	0.855 52	0.579 61	0.589 56	0.735 55	0.484 41	0.588 49	0.856 20	0.634 35	0.571 43	0.298 53	0.500 54	1.000 1	0.824 38	0.818 46	0.702 41	0.935 57	0.545 50
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)
Region-18class	0.497 61	0.250 68	0.902 43	0.689 52	0.540 59	0.747 51	0.276 57	0.610 46	0.268 67	0.489 58	0.348 59	0.000 67	0.243 67	0.220 66	0.663 54	0.814 49	0.459 63	0.928 58	0.496 56

3D-BoNet	0.687 50	1.000 1	0.887 49	0.836 23	0.587 57	0.643 64	0.550 35	0.620 45	0.724 42	0.522 56	0.501 54	0.243 56	0.512 52	1.000 1	0.751 50	0.807 51	0.661 50	0.909 59	0.612 43
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
Occipital-SCS	0.688 49	1.000 1	0.913 41	0.730 49	0.737 39	0.743 54	0.442 45	0.855 11	0.655 49	0.546 52	0.546 48	0.263 55	0.508 53	0.889 44	0.568 59	0.771 58	0.705 40	0.889 60	0.625 38

PCJC	0.684 52	1.000 1	0.895 48	0.757 45	0.659 50	0.862 31	0.189 63	0.739 32	0.606 53	0.712 24	0.581 42	0.515 42	0.650 23	0.857 46	0.357 64	0.785 56	0.631 52	0.889 60	0.635 37

3D-BEVIS	0.401 66	0.667 64	0.687 64	0.419 67	0.137 68	0.587 66	0.188 64	0.235 63	0.359 65	0.211 67	0.093 68	0.080 64	0.311 63	0.571 61	0.382 61	0.754 60	0.300 67	0.874 62	0.357 61
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
Sem_Recon_ins	0.484 62	0.764 62	0.608 67	0.470 64	0.521 60	0.637 65	0.311 52	0.218 64	0.348 66	0.365 65	0.223 64	0.222 57	0.258 65	0.629 59	0.734 51	0.596 67	0.509 58	0.858 63	0.444 57

tmp	0.474 63	1.000 1	0.727 60	0.433 66	0.481 63	0.673 62	0.022 68	0.380 60	0.517 61	0.436 63	0.338 61	0.128 61	0.343 62	0.429 63	0.291 66	0.728 62	0.473 60	0.833 64	0.300 65

SemRegionNet-20cls	0.470 64	1.000 1	0.727 60	0.447 65	0.481 62	0.678 61	0.024 67	0.380 60	0.518 60	0.440 62	0.339 60	0.128 61	0.350 61	0.429 63	0.212 67	0.711 63	0.465 62	0.833 64	0.290 66

Sgpn_scannet	0.390 67	0.556 65	0.636 66	0.493 63	0.353 66	0.539 67	0.271 59	0.160 67	0.450 64	0.359 66	0.178 66	0.146 60	0.250 66	0.143 67	0.347 65	0.698 64	0.436 66	0.667 66	0.331 64

ASIS	0.422 65	0.333 67	0.707 63	0.676 53	0.401 65	0.650 63	0.350 50	0.177 66	0.594 55	0.376 64	0.202 65	0.077 65	0.404 59	0.571 61	0.197 68	0.674 65	0.447 64	0.500 67	0.260 67

MaskRCNN 2d->3d Proj	0.261 68	0.903 61	0.081 68	0.008 68	0.233 67	0.175 68	0.280 55	0.106 68	0.150 68	0.203 68	0.175 67	0.480 44	0.218 68	0.143 67	0.542 60	0.404 68	0.153 68	0.393 68	0.049 68

3D Semantic Instance Benchmark