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

ExtMask3D	0.598 5	0.852 15	0.692 6	0.433 28	0.461 6	0.791 1	0.264 12	0.488 33	0.493 1	0.508 3	0.528 12	0.594 10	0.706 6	0.791 8	0.483 5	0.734 9	0.595 3	0.911 17	0.437 3

MG-Former	0.587 8	0.852 15	0.639 13	0.454 20	0.393 18	0.758 11	0.338 2	0.572 16	0.480 2	0.527 2	0.491 19	0.671 4	0.527 20	0.867 1	0.485 4	0.601 28	0.590 6	0.938 9	0.390 10

UniPerception	0.588 7	0.963 3	0.667 8	0.493 13	0.472 5	0.750 12	0.229 16	0.528 26	0.468 3	0.498 10	0.542 6	0.643 6	0.530 19	0.661 35	0.463 14	0.695 17	0.599 2	0.972 1	0.420 6

EV3D	0.615 4	0.946 5	0.652 11	0.555 5	0.433 11	0.773 3	0.271 11	0.604 8	0.447 4	0.506 4	0.544 5	0.698 2	0.716 2	0.775 14	0.480 7	0.747 5	0.572 11	0.925 11	0.435 5

Spherical Mask(CtoF)	0.616 3	0.946 5	0.654 10	0.555 5	0.434 10	0.769 7	0.271 10	0.604 8	0.447 4	0.505 5	0.549 2	0.698 2	0.716 2	0.775 14	0.480 7	0.747 5	0.575 9	0.925 11	0.436 4

InsSSM	0.586 9	1.000 1	0.593 18	0.440 24	0.480 3	0.771 5	0.345 1	0.437 37	0.444 6	0.495 11	0.548 4	0.579 13	0.621 9	0.720 26	0.409 20	0.712 11	0.593 4	0.960 4	0.395 8

PointRel	0.622 1	0.926 8	0.710 2	0.541 8	0.502 2	0.772 4	0.314 4	0.598 11	0.425 7	0.504 7	0.565 1	0.650 5	0.716 2	0.809 7	0.476 10	0.747 4	0.618 1	0.963 3	0.364 18
: Relation3D (PointRel): Enhancing Relation Modeling for Point Cloud Instance Segmentation.
SIM3D	0.617 2	0.952 4	0.629 14	0.539 9	0.426 13	0.768 8	0.302 6	0.681 2	0.425 8	0.473 13	0.511 13	0.701 1	0.717 1	0.821 6	0.467 13	0.774 1	0.559 13	0.914 15	0.448 2

MAFT	0.596 6	0.889 13	0.721 1	0.448 21	0.460 7	0.768 9	0.251 13	0.558 21	0.408 9	0.504 6	0.539 7	0.616 8	0.618 10	0.858 3	0.482 6	0.684 18	0.551 15	0.931 10	0.450 1

TST3D	0.569 13	0.778 23	0.675 7	0.598 2	0.451 8	0.727 19	0.280 8	0.476 36	0.395 10	0.472 14	0.457 25	0.583 11	0.580 14	0.777 11	0.462 16	0.735 8	0.547 17	0.919 14	0.333 25
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
ISBNet	0.559 16	0.939 7	0.655 9	0.383 37	0.426 14	0.763 10	0.180 18	0.534 25	0.386 11	0.499 8	0.509 15	0.621 7	0.427 35	0.704 30	0.467 12	0.649 21	0.571 12	0.948 7	0.401 7
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
Mask3D	0.566 14	0.926 8	0.597 17	0.408 31	0.420 15	0.737 15	0.239 14	0.598 11	0.386 12	0.458 17	0.549 2	0.568 16	0.716 2	0.601 41	0.480 7	0.646 22	0.575 9	0.922 13	0.364 17
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
PBNet	0.573 12	0.926 8	0.575 23	0.619 1	0.472 4	0.736 16	0.239 15	0.487 34	0.383 13	0.459 16	0.506 16	0.533 19	0.585 12	0.767 16	0.404 21	0.717 10	0.559 14	0.969 2	0.381 14
W.Zhao, Y.Yan, C.Yang, J.Ye,X.Yang,K.Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
Competitor-SPFormer	0.580 11	0.721 31	0.705 3	0.593 3	0.444 9	0.786 2	0.286 7	0.564 19	0.376 14	0.498 9	0.534 10	0.546 18	0.390 41	0.785 10	0.577 1	0.708 15	0.579 8	0.954 6	0.388 11

Queryformer	0.583 10	0.926 8	0.702 4	0.393 34	0.504 1	0.733 18	0.276 9	0.527 27	0.373 15	0.479 12	0.534 9	0.533 20	0.697 7	0.720 27	0.436 18	0.745 7	0.592 5	0.958 5	0.363 19

OneFormer3D	0.566 14	0.781 22	0.697 5	0.562 4	0.431 12	0.770 6	0.331 3	0.400 43	0.373 16	0.529 1	0.504 17	0.568 15	0.475 25	0.732 24	0.470 11	0.762 2	0.550 16	0.871 32	0.379 15
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
TD3D	0.489 25	0.852 15	0.511 37	0.434 26	0.322 31	0.735 17	0.101 26	0.512 29	0.355 17	0.349 24	0.468 23	0.283 32	0.514 22	0.676 34	0.268 36	0.671 19	0.510 20	0.908 18	0.329 27
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
SPFormer	0.549 18	0.745 26	0.640 12	0.484 14	0.395 17	0.739 14	0.311 5	0.566 18	0.335 18	0.468 15	0.492 18	0.555 17	0.478 24	0.747 21	0.436 17	0.712 12	0.540 18	0.893 24	0.343 24
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
GraphCut	0.552 17	1.000 1	0.611 16	0.438 25	0.392 19	0.714 20	0.139 21	0.598 13	0.327 19	0.389 20	0.510 14	0.598 9	0.427 36	0.754 19	0.463 15	0.761 3	0.588 7	0.903 20	0.329 26

DualGroup	0.469 28	0.815 19	0.552 26	0.398 32	0.374 22	0.683 26	0.130 22	0.539 24	0.310 20	0.327 27	0.407 31	0.276 33	0.447 29	0.535 45	0.342 25	0.659 20	0.455 28	0.900 23	0.301 31

DKNet	0.532 19	0.815 19	0.624 15	0.517 10	0.377 21	0.749 13	0.107 23	0.509 30	0.304 21	0.437 18	0.475 20	0.581 12	0.539 17	0.775 13	0.339 26	0.640 24	0.506 21	0.901 21	0.385 13
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
SoftGroup	0.504 23	0.667 40	0.579 20	0.372 39	0.381 20	0.694 23	0.072 29	0.677 3	0.303 22	0.387 21	0.531 11	0.319 28	0.582 13	0.754 18	0.318 27	0.643 23	0.492 24	0.907 19	0.388 12
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral]
OccuSeg+instance	0.486 26	0.802 21	0.536 30	0.428 29	0.369 23	0.702 22	0.205 17	0.331 51	0.301 23	0.379 22	0.474 21	0.327 25	0.437 30	0.862 2	0.485 3	0.601 29	0.394 38	0.846 42	0.273 39
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
OSIS	0.392 41	0.778 23	0.530 31	0.220 54	0.278 38	0.567 48	0.083 28	0.330 52	0.299 24	0.270 35	0.310 44	0.143 51	0.260 51	0.624 39	0.277 34	0.568 38	0.361 43	0.865 34	0.301 30

SSEN	0.384 43	0.852 15	0.494 43	0.192 55	0.226 48	0.648 35	0.022 47	0.398 44	0.299 25	0.277 33	0.317 43	0.231 41	0.194 58	0.514 48	0.196 45	0.586 33	0.444 29	0.843 43	0.184 51
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv
SoftGroup++	0.513 21	0.704 33	0.578 22	0.398 33	0.363 27	0.704 21	0.061 33	0.647 5	0.297 26	0.378 23	0.537 8	0.343 24	0.614 11	0.828 5	0.295 31	0.710 14	0.505 23	0.875 31	0.394 9

SSEC	0.465 29	0.667 40	0.578 21	0.502 11	0.362 28	0.641 37	0.035 42	0.605 7	0.291 27	0.323 28	0.451 26	0.296 30	0.417 39	0.677 33	0.245 40	0.501 49	0.506 22	0.900 22	0.366 16

TopoSeg	0.479 27	0.704 33	0.564 24	0.467 18	0.366 25	0.633 38	0.068 30	0.554 22	0.262 28	0.328 26	0.447 28	0.323 26	0.534 18	0.722 25	0.288 33	0.614 26	0.482 25	0.912 16	0.358 22

IPCA-Inst	0.520 20	0.889 13	0.551 27	0.548 7	0.418 16	0.665 30	0.064 32	0.585 14	0.260 29	0.277 34	0.471 22	0.500 21	0.644 8	0.785 9	0.369 22	0.591 31	0.511 19	0.878 29	0.362 20

Mask-Group	0.434 33	0.778 23	0.516 35	0.471 17	0.330 30	0.658 31	0.029 44	0.526 28	0.249 30	0.256 37	0.400 32	0.309 29	0.384 44	0.296 61	0.368 23	0.575 35	0.425 33	0.877 30	0.362 21
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022
DANCENET	0.504 23	0.926 8	0.579 19	0.472 16	0.367 24	0.626 40	0.165 20	0.432 38	0.221 31	0.408 19	0.449 27	0.411 22	0.564 15	0.746 22	0.421 19	0.707 16	0.438 30	0.846 40	0.288 36

SphereSeg	0.357 47	0.651 45	0.411 49	0.345 42	0.264 42	0.630 39	0.059 34	0.289 58	0.212 32	0.240 40	0.336 40	0.158 50	0.305 48	0.557 42	0.159 54	0.455 55	0.341 48	0.726 55	0.294 32

3D-MPA	0.355 48	0.457 60	0.484 44	0.299 47	0.277 39	0.591 47	0.047 40	0.332 49	0.212 33	0.217 45	0.278 46	0.193 46	0.413 40	0.410 55	0.195 46	0.574 37	0.352 44	0.849 38	0.213 48
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020
INS-Conv-instance	0.435 32	0.716 32	0.495 42	0.355 41	0.331 29	0.689 25	0.102 25	0.394 45	0.208 34	0.280 32	0.395 33	0.250 36	0.544 16	0.741 23	0.309 29	0.536 45	0.391 39	0.842 45	0.258 43

SSTNet	0.506 22	0.738 29	0.549 28	0.497 12	0.316 32	0.693 24	0.178 19	0.377 46	0.198 35	0.330 25	0.463 24	0.576 14	0.515 21	0.857 4	0.494 2	0.637 25	0.457 27	0.943 8	0.290 35
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
HAIS	0.457 30	0.704 33	0.561 25	0.457 19	0.364 26	0.673 27	0.046 41	0.547 23	0.194 36	0.308 29	0.426 29	0.288 31	0.454 28	0.711 28	0.262 37	0.563 39	0.434 32	0.889 26	0.344 23
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
DD-UNet+Group	0.436 31	0.630 48	0.508 40	0.480 15	0.310 34	0.624 42	0.065 31	0.638 6	0.174 37	0.256 38	0.384 35	0.194 45	0.428 33	0.759 17	0.289 32	0.574 36	0.400 36	0.849 39	0.291 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
Dyco3D	0.395 40	0.642 46	0.518 34	0.447 22	0.259 43	0.666 29	0.050 38	0.251 61	0.166 38	0.231 42	0.362 37	0.232 40	0.331 47	0.535 44	0.229 41	0.587 32	0.438 31	0.850 37	0.317 28
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
NeuralBF	0.353 49	0.593 50	0.511 38	0.375 38	0.264 41	0.597 45	0.008 54	0.332 50	0.160 39	0.229 43	0.274 49	0.000 72	0.206 55	0.678 32	0.155 55	0.485 51	0.422 34	0.816 48	0.254 44
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
Mask3D_evaluation	0.382 44	0.593 50	0.520 33	0.390 35	0.314 33	0.600 44	0.018 51	0.287 59	0.151 40	0.281 31	0.387 34	0.169 49	0.429 32	0.654 37	0.172 51	0.578 34	0.384 40	0.670 58	0.278 38

DENet	0.413 36	0.741 27	0.520 32	0.237 52	0.284 36	0.523 51	0.097 27	0.691 1	0.138 41	0.209 51	0.229 53	0.238 39	0.390 42	0.707 29	0.310 28	0.448 56	0.470 26	0.892 25	0.310 29

PCJC	0.375 45	0.704 33	0.542 29	0.284 49	0.197 52	0.649 34	0.006 56	0.426 39	0.138 42	0.242 39	0.304 45	0.183 48	0.388 43	0.629 38	0.141 58	0.546 43	0.344 46	0.738 53	0.283 37

Sparse R-CNN	0.292 54	0.704 33	0.213 64	0.153 58	0.154 55	0.551 50	0.053 35	0.212 62	0.132 43	0.174 56	0.274 48	0.070 57	0.363 45	0.441 54	0.176 50	0.424 58	0.234 59	0.758 51	0.161 56

RPGN	0.428 35	0.630 48	0.508 39	0.367 40	0.249 44	0.658 32	0.016 52	0.673 4	0.131 44	0.234 41	0.383 36	0.270 34	0.434 31	0.748 20	0.274 35	0.609 27	0.406 35	0.842 44	0.267 42
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
AOIA	0.387 42	0.704 33	0.515 36	0.385 36	0.225 49	0.669 28	0.005 59	0.482 35	0.126 45	0.181 54	0.269 50	0.221 42	0.426 37	0.478 51	0.218 43	0.592 30	0.371 41	0.851 36	0.242 45

Box2Mask	0.433 34	0.741 27	0.463 47	0.433 27	0.283 37	0.625 41	0.103 24	0.298 56	0.125 46	0.260 36	0.424 30	0.322 27	0.472 26	0.701 31	0.363 24	0.711 13	0.309 55	0.882 27	0.272 41
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022
PointGroup	0.407 37	0.639 47	0.496 41	0.415 30	0.243 46	0.645 36	0.021 49	0.570 17	0.114 47	0.211 49	0.359 38	0.217 43	0.428 34	0.660 36	0.256 38	0.562 40	0.341 47	0.860 35	0.291 33
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	0.396 39	0.667 40	0.467 45	0.446 23	0.243 45	0.624 43	0.022 48	0.577 15	0.106 48	0.219 44	0.340 39	0.239 38	0.487 23	0.475 52	0.225 42	0.541 44	0.350 45	0.818 47	0.273 40
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
GICN	0.341 51	0.580 52	0.371 52	0.344 43	0.198 51	0.469 54	0.052 36	0.564 20	0.093 49	0.212 47	0.212 55	0.127 53	0.347 46	0.537 43	0.206 44	0.525 46	0.329 50	0.729 54	0.241 46

ClickSeg_Instance	0.366 46	0.654 44	0.375 51	0.184 56	0.302 35	0.592 46	0.050 39	0.300 55	0.093 50	0.283 30	0.277 47	0.249 37	0.426 38	0.615 40	0.299 30	0.504 48	0.367 42	0.832 46	0.191 50

CSC-Pretrained	0.405 38	0.738 29	0.465 46	0.331 45	0.205 50	0.655 33	0.051 37	0.601 10	0.092 51	0.211 50	0.329 41	0.198 44	0.459 27	0.775 12	0.195 47	0.524 47	0.400 37	0.878 28	0.184 52

MASC	0.254 57	0.463 59	0.249 63	0.113 59	0.167 54	0.412 59	0.000 64	0.374 47	0.073 52	0.173 57	0.243 51	0.130 52	0.228 54	0.368 57	0.160 53	0.356 60	0.208 60	0.711 57	0.136 58
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
3D-BoNet	0.253 58	0.519 55	0.324 58	0.251 51	0.137 58	0.345 64	0.031 43	0.419 41	0.069 53	0.162 58	0.131 60	0.052 59	0.202 57	0.338 59	0.147 57	0.301 65	0.303 56	0.651 60	0.178 53
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
SPG_WSIS	0.251 59	0.380 62	0.274 61	0.289 48	0.144 56	0.413 58	0.000 65	0.311 53	0.065 54	0.113 61	0.130 61	0.029 64	0.204 56	0.388 56	0.108 61	0.459 54	0.311 53	0.769 50	0.127 59

SALoss-ResNet	0.262 56	0.667 40	0.335 55	0.067 65	0.123 59	0.427 57	0.022 46	0.280 60	0.058 55	0.216 46	0.211 56	0.039 61	0.142 61	0.519 47	0.106 62	0.338 62	0.310 54	0.721 56	0.138 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)
Occipital-SCS	0.320 53	0.679 39	0.352 54	0.334 44	0.229 47	0.436 55	0.025 45	0.412 42	0.058 56	0.161 59	0.240 52	0.085 55	0.262 50	0.496 50	0.187 48	0.467 53	0.328 51	0.775 49	0.231 47

MTML	0.282 55	0.577 53	0.380 50	0.182 57	0.107 61	0.430 56	0.001 62	0.422 40	0.057 57	0.179 55	0.162 58	0.070 58	0.229 53	0.511 49	0.161 52	0.491 50	0.313 52	0.650 61	0.162 54
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
RWSeg	0.348 50	0.475 57	0.456 48	0.320 46	0.275 40	0.476 53	0.020 50	0.491 32	0.056 58	0.212 48	0.320 42	0.261 35	0.302 49	0.520 46	0.182 49	0.557 41	0.285 57	0.867 33	0.197 49

R-PointNet	0.158 64	0.356 63	0.173 67	0.113 60	0.140 57	0.359 60	0.012 53	0.023 70	0.039 59	0.134 60	0.123 62	0.008 68	0.089 64	0.149 66	0.117 60	0.221 67	0.128 67	0.563 63	0.094 61

One_Thing_One_Click	0.326 52	0.472 58	0.361 53	0.232 53	0.183 53	0.555 49	0.000 65	0.498 31	0.038 60	0.195 52	0.226 54	0.362 23	0.168 59	0.469 53	0.251 39	0.553 42	0.335 49	0.846 41	0.117 60
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
3D-BEVIS	0.117 67	0.250 67	0.308 60	0.020 71	0.009 73	0.269 69	0.006 57	0.008 71	0.029 61	0.037 71	0.014 72	0.003 70	0.036 69	0.147 67	0.042 66	0.381 59	0.118 68	0.362 67	0.069 66
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
PanopticFusion-inst	0.214 61	0.250 67	0.330 57	0.275 50	0.103 62	0.228 70	0.000 65	0.345 48	0.024 62	0.088 63	0.203 57	0.186 47	0.167 60	0.367 58	0.125 59	0.221 68	0.112 70	0.666 59	0.162 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)
SemRegionNet-20cls	0.121 66	0.296 65	0.203 65	0.071 63	0.058 65	0.349 62	0.000 65	0.150 63	0.019 63	0.054 68	0.034 69	0.017 67	0.052 66	0.042 70	0.013 71	0.209 69	0.183 61	0.371 66	0.057 68

SegGroup_ins	0.246 60	0.556 54	0.335 56	0.062 67	0.115 60	0.490 52	0.000 65	0.297 57	0.018 64	0.186 53	0.142 59	0.083 56	0.233 52	0.216 63	0.153 56	0.469 52	0.251 58	0.744 52	0.083 63
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
tmp	0.113 69	0.333 64	0.151 70	0.056 68	0.053 66	0.344 65	0.000 65	0.105 66	0.016 65	0.049 69	0.035 68	0.020 66	0.053 65	0.048 69	0.013 70	0.183 71	0.173 63	0.344 69	0.054 69

UNet-backbone	0.161 62	0.519 55	0.259 62	0.084 61	0.059 64	0.325 66	0.002 60	0.093 67	0.009 66	0.077 65	0.064 64	0.045 60	0.044 68	0.161 65	0.045 64	0.331 63	0.180 62	0.566 62	0.033 72

3D-SIS	0.161 62	0.407 61	0.155 69	0.068 64	0.043 68	0.346 63	0.001 61	0.134 64	0.005 67	0.088 62	0.106 63	0.037 62	0.135 63	0.321 60	0.028 68	0.339 61	0.116 69	0.466 65	0.093 62
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
Sem_Recon_ins	0.098 70	0.295 66	0.187 66	0.015 72	0.036 69	0.213 71	0.005 58	0.038 69	0.003 68	0.056 67	0.037 67	0.036 63	0.015 71	0.051 68	0.044 65	0.209 70	0.098 71	0.354 68	0.071 65

Region-18class	0.146 65	0.175 71	0.321 59	0.080 62	0.062 63	0.357 61	0.000 65	0.307 54	0.002 69	0.066 66	0.044 66	0.000 72	0.018 70	0.036 71	0.054 63	0.447 57	0.133 65	0.472 64	0.060 67

ASIS	0.085 71	0.037 72	0.080 72	0.066 66	0.047 67	0.282 68	0.000 65	0.052 68	0.002 70	0.047 70	0.026 70	0.001 71	0.046 67	0.194 64	0.031 67	0.264 66	0.140 64	0.167 71	0.047 71

Hier3D	0.117 67	0.222 69	0.161 68	0.054 69	0.027 70	0.289 67	0.000 65	0.124 65	0.001 71	0.079 64	0.061 65	0.027 65	0.141 62	0.240 62	0.005 72	0.310 64	0.129 66	0.153 72	0.081 64
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
Sgpn_scannet	0.049 72	0.023 73	0.134 71	0.031 70	0.013 72	0.144 72	0.006 55	0.008 72	0.000 72	0.028 72	0.017 71	0.003 69	0.009 73	0.000 72	0.021 69	0.122 72	0.095 72	0.175 70	0.054 70

MaskRCNN 2d->3d Proj	0.022 73	0.185 70	0.000 73	0.000 73	0.015 71	0.000 73	0.000 63	0.006 73	0.000 72	0.010 73	0.006 73	0.107 54	0.012 72	0.000 72	0.002 73	0.027 73	0.004 73	0.022 73	0.001 73

3D Semantic Instance Benchmark