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

InsSSM	0.586 11	1.000 1	0.593 21	0.440 26	0.480 5	0.771 8	0.345 1	0.437 40	0.444 7	0.495 14	0.548 5	0.579 16	0.621 11	0.720 28	0.409 23	0.712 11	0.593 6	0.960 5	0.395 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	0.587 10	0.852 17	0.639 15	0.454 22	0.393 21	0.758 14	0.338 2	0.572 16	0.480 3	0.527 3	0.491 22	0.671 6	0.527 22	0.867 1	0.485 6	0.601 31	0.590 8	0.938 12	0.390 12

OneFormer3D	0.566 17	0.781 24	0.697 6	0.562 5	0.431 15	0.770 9	0.331 3	0.400 46	0.373 18	0.529 2	0.504 20	0.568 18	0.475 29	0.732 26	0.470 13	0.762 2	0.550 19	0.871 35	0.379 17
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
PointRel	0.622 1	0.926 8	0.710 3	0.541 10	0.502 2	0.772 7	0.314 4	0.598 11	0.425 8	0.504 10	0.565 2	0.650 7	0.716 2	0.809 7	0.476 12	0.747 4	0.618 2	0.963 4	0.364 20
: Relation3D (PointRel): Enhancing Relation Modeling for Point Cloud Instance Segmentation.
SPFormer	0.549 21	0.745 28	0.640 14	0.484 16	0.395 20	0.739 17	0.311 5	0.566 18	0.335 21	0.468 18	0.492 21	0.555 20	0.478 28	0.747 23	0.436 20	0.712 12	0.540 21	0.893 27	0.343 27
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
SIM3D	0.617 3	0.952 4	0.629 17	0.539 11	0.426 16	0.768 11	0.302 6	0.681 2	0.425 9	0.473 16	0.511 16	0.701 2	0.717 1	0.821 6	0.467 15	0.774 1	0.559 15	0.914 18	0.448 3

Competitor-MAFT	0.618 2	0.866 16	0.724 1	0.628 1	0.484 4	0.803 2	0.300 7	0.509 32	0.496 1	0.539 1	0.547 6	0.703 1	0.668 8	0.708 32	0.463 17	0.708 16	0.595 4	0.959 6	0.418 8

DCD	0.614 6	0.892 13	0.633 16	0.434 28	0.495 3	0.810 1	0.292 8	0.501 33	0.408 10	0.525 4	0.582 1	0.688 5	0.625 10	0.801 8	0.608 1	0.672 20	0.649 1	0.965 3	0.476 1

Competitor-SPFormer	0.580 14	0.721 35	0.705 4	0.593 4	0.444 12	0.786 5	0.286 9	0.564 19	0.376 16	0.498 12	0.534 12	0.546 21	0.390 45	0.785 12	0.577 2	0.708 15	0.579 10	0.954 8	0.388 13

TST3D	0.569 16	0.778 25	0.675 9	0.598 3	0.451 11	0.727 22	0.280 10	0.476 39	0.395 12	0.472 17	0.457 28	0.583 14	0.580 16	0.777 13	0.462 19	0.735 8	0.547 20	0.919 17	0.333 28
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
Queryformer	0.583 12	0.926 8	0.702 5	0.393 37	0.504 1	0.733 21	0.276 11	0.527 28	0.373 17	0.479 15	0.534 11	0.533 23	0.697 7	0.720 29	0.436 21	0.745 7	0.592 7	0.958 7	0.363 21

Spherical Mask(CtoF)	0.616 4	0.946 5	0.654 12	0.555 6	0.434 13	0.769 10	0.271 12	0.604 8	0.447 5	0.505 8	0.549 3	0.698 3	0.716 2	0.775 16	0.480 9	0.747 5	0.575 11	0.925 14	0.436 5

EV3D	0.615 5	0.946 5	0.652 13	0.555 6	0.433 14	0.773 6	0.271 13	0.604 8	0.447 5	0.506 7	0.544 7	0.698 3	0.716 2	0.775 16	0.480 9	0.747 5	0.572 13	0.925 14	0.435 6

ExtMask3D	0.598 7	0.852 17	0.692 7	0.433 31	0.461 8	0.791 4	0.264 14	0.488 36	0.493 2	0.508 6	0.528 15	0.594 13	0.706 6	0.791 10	0.483 7	0.734 9	0.595 5	0.911 20	0.437 4

KmaxOneFormerNet	0.581 13	0.745 28	0.692 8	0.551 8	0.458 10	0.798 3	0.264 15	0.531 26	0.369 19	0.513 5	0.531 14	0.632 9	0.494 25	0.798 9	0.567 3	0.648 24	0.558 17	0.950 9	0.362 22

MAFT	0.596 8	0.889 14	0.721 2	0.448 23	0.460 9	0.768 12	0.251 16	0.558 21	0.408 11	0.504 9	0.539 9	0.616 11	0.618 12	0.858 3	0.482 8	0.684 19	0.551 18	0.931 13	0.450 2

Mask3D	0.566 17	0.926 8	0.597 20	0.408 34	0.420 18	0.737 18	0.239 17	0.598 11	0.386 14	0.458 20	0.549 3	0.568 19	0.716 2	0.601 45	0.480 9	0.646 25	0.575 11	0.922 16	0.364 19
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
PBNet	0.573 15	0.926 8	0.575 27	0.619 2	0.472 6	0.736 19	0.239 18	0.487 37	0.383 15	0.459 19	0.506 19	0.533 22	0.585 14	0.767 18	0.404 24	0.717 10	0.559 16	0.969 2	0.381 16
Weiguang Zhao, Yuyao Yan, Chaolong Yang, Jianan Ye, Xi Yang, Kaizhu Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
UniPerception	0.588 9	0.963 3	0.667 10	0.493 15	0.472 7	0.750 15	0.229 19	0.528 27	0.468 4	0.498 13	0.542 8	0.643 8	0.530 21	0.661 39	0.463 16	0.695 18	0.599 3	0.972 1	0.420 7

OccuSeg+instance	0.486 29	0.802 23	0.536 34	0.428 32	0.369 26	0.702 25	0.205 20	0.331 55	0.301 27	0.379 25	0.474 24	0.327 29	0.437 34	0.862 2	0.485 5	0.601 32	0.394 41	0.846 45	0.273 43
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
ISBNet	0.559 19	0.939 7	0.655 11	0.383 40	0.426 17	0.763 13	0.180 21	0.534 25	0.386 13	0.499 11	0.509 18	0.621 10	0.427 39	0.704 34	0.467 14	0.649 23	0.571 14	0.948 10	0.401 9
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
SSTNet	0.506 25	0.738 32	0.549 32	0.497 14	0.316 36	0.693 27	0.178 22	0.377 50	0.198 39	0.330 29	0.463 27	0.576 17	0.515 23	0.857 4	0.494 4	0.637 28	0.457 30	0.943 11	0.290 39
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
DANCENET	0.504 26	0.926 8	0.579 23	0.472 18	0.367 27	0.626 43	0.165 23	0.432 41	0.221 35	0.408 22	0.449 30	0.411 26	0.564 17	0.746 24	0.421 22	0.707 17	0.438 33	0.846 43	0.288 40

ODIN - Ins	0.463 33	0.738 32	0.589 22	0.344 46	0.358 32	0.560 52	0.139 24	0.393 49	0.331 22	0.373 27	0.392 37	0.496 25	0.493 26	0.709 31	0.377 25	0.599 33	0.359 47	0.752 55	0.332 29
Ayush Jain, Pushkal Katara, Nikolaos Gkanatsios, Adam W. Harley, Gabriel Sarch, Kriti Aggarwal, Vishrav Chaudhary, Katerina Fragkiadaki: ODIN: A Single Model for 2D and 3D Segmentation. CVPR 2024
GraphCut	0.552 20	1.000 1	0.611 19	0.438 27	0.392 22	0.714 23	0.139 25	0.598 13	0.327 23	0.389 23	0.510 17	0.598 12	0.427 40	0.754 21	0.463 18	0.761 3	0.588 9	0.903 23	0.329 30

DualGroup	0.469 31	0.815 21	0.552 30	0.398 35	0.374 25	0.683 29	0.130 26	0.539 24	0.310 24	0.327 31	0.407 34	0.276 37	0.447 33	0.535 49	0.342 29	0.659 22	0.455 31	0.900 26	0.301 35

DKNet	0.532 22	0.815 21	0.624 18	0.517 12	0.377 24	0.749 16	0.107 27	0.509 31	0.304 25	0.437 21	0.475 23	0.581 15	0.539 19	0.775 15	0.339 30	0.640 27	0.506 24	0.901 24	0.385 15
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
Box2Mask	0.433 38	0.741 30	0.463 51	0.433 30	0.283 41	0.625 44	0.103 28	0.298 60	0.125 50	0.260 40	0.424 33	0.322 31	0.472 30	0.701 35	0.363 28	0.711 13	0.309 59	0.882 30	0.272 45
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022
INS-Conv-instance	0.435 36	0.716 36	0.495 46	0.355 44	0.331 33	0.689 28	0.102 29	0.394 48	0.208 38	0.280 36	0.395 36	0.250 40	0.544 18	0.741 25	0.309 33	0.536 49	0.391 42	0.842 48	0.258 47

TD3D	0.489 28	0.852 17	0.511 41	0.434 29	0.322 35	0.735 20	0.101 30	0.512 30	0.355 20	0.349 28	0.468 26	0.283 36	0.514 24	0.676 38	0.268 40	0.671 21	0.510 23	0.908 21	0.329 31
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
DENet	0.413 40	0.741 30	0.520 36	0.237 56	0.284 40	0.523 55	0.097 31	0.691 1	0.138 45	0.209 55	0.229 57	0.238 43	0.390 46	0.707 33	0.310 32	0.448 60	0.470 29	0.892 28	0.310 33

OSIS	0.392 45	0.778 25	0.530 35	0.220 58	0.278 42	0.567 51	0.083 32	0.330 56	0.299 28	0.270 39	0.310 48	0.143 55	0.260 55	0.624 43	0.277 38	0.568 42	0.361 46	0.865 37	0.301 34

SoftGroup	0.504 26	0.667 44	0.579 24	0.372 42	0.381 23	0.694 26	0.072 33	0.677 3	0.303 26	0.387 24	0.531 13	0.319 32	0.582 15	0.754 20	0.318 31	0.643 26	0.492 27	0.907 22	0.388 14
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral]
TopoSeg	0.479 30	0.704 37	0.564 28	0.467 20	0.366 28	0.633 41	0.068 34	0.554 22	0.262 32	0.328 30	0.447 31	0.323 30	0.534 20	0.722 27	0.288 37	0.614 29	0.482 28	0.912 19	0.358 25

DD-UNet+Group	0.436 35	0.630 52	0.508 44	0.480 17	0.310 38	0.624 45	0.065 35	0.638 6	0.174 41	0.256 42	0.384 39	0.194 49	0.428 37	0.759 19	0.289 36	0.574 40	0.400 39	0.849 42	0.291 38
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
IPCA-Inst	0.520 23	0.889 14	0.551 31	0.548 9	0.418 19	0.665 33	0.064 36	0.585 14	0.260 33	0.277 38	0.471 25	0.500 24	0.644 9	0.785 11	0.369 26	0.591 35	0.511 22	0.878 32	0.362 23

SoftGroup++	0.513 24	0.704 37	0.578 26	0.398 36	0.363 30	0.704 24	0.061 37	0.647 5	0.297 30	0.378 26	0.537 10	0.343 28	0.614 13	0.828 5	0.295 35	0.710 14	0.505 26	0.875 34	0.394 11

SphereSeg	0.357 51	0.651 49	0.411 53	0.345 45	0.264 46	0.630 42	0.059 38	0.289 62	0.212 36	0.240 44	0.336 44	0.158 54	0.305 52	0.557 46	0.159 58	0.455 59	0.341 52	0.726 59	0.294 36

Sparse R-CNN	0.292 58	0.704 37	0.213 68	0.153 62	0.154 59	0.551 54	0.053 39	0.212 66	0.132 47	0.174 60	0.274 52	0.070 61	0.363 49	0.441 58	0.176 54	0.424 62	0.234 63	0.758 54	0.161 60

GICN	0.341 55	0.580 56	0.371 56	0.344 47	0.198 55	0.469 58	0.052 40	0.564 20	0.093 53	0.212 51	0.212 59	0.127 57	0.347 50	0.537 47	0.206 48	0.525 50	0.329 54	0.729 58	0.241 50

CSC-Pretrained	0.405 42	0.738 32	0.465 50	0.331 49	0.205 54	0.655 36	0.051 41	0.601 10	0.092 55	0.211 54	0.329 45	0.198 48	0.459 31	0.775 14	0.195 51	0.524 51	0.400 40	0.878 31	0.184 56

Dyco3D	0.395 44	0.642 50	0.518 38	0.447 24	0.259 47	0.666 32	0.050 42	0.251 65	0.166 42	0.231 46	0.362 41	0.232 44	0.331 51	0.535 48	0.229 45	0.587 36	0.438 34	0.850 40	0.317 32
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
ClickSeg_Instance	0.366 50	0.654 48	0.375 55	0.184 60	0.302 39	0.592 49	0.050 43	0.300 59	0.093 54	0.283 34	0.277 51	0.249 41	0.426 42	0.615 44	0.299 34	0.504 52	0.367 45	0.832 49	0.191 54

3D-MPA	0.355 52	0.457 64	0.484 48	0.299 51	0.277 43	0.591 50	0.047 44	0.332 53	0.212 37	0.217 49	0.278 50	0.193 50	0.413 44	0.410 59	0.195 50	0.574 41	0.352 48	0.849 41	0.213 52
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020
HAIS	0.457 34	0.704 37	0.561 29	0.457 21	0.364 29	0.673 30	0.046 45	0.547 23	0.194 40	0.308 33	0.426 32	0.288 35	0.454 32	0.711 30	0.262 41	0.563 43	0.434 35	0.889 29	0.344 26
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
SSEC	0.465 32	0.667 44	0.578 25	0.502 13	0.362 31	0.641 40	0.035 46	0.605 7	0.291 31	0.323 32	0.451 29	0.296 34	0.417 43	0.677 37	0.245 44	0.501 53	0.506 25	0.900 25	0.366 18

3D-BoNet	0.253 62	0.519 59	0.324 62	0.251 55	0.137 62	0.345 68	0.031 47	0.419 44	0.069 57	0.162 62	0.131 64	0.052 63	0.202 61	0.338 63	0.147 61	0.301 69	0.303 60	0.651 64	0.178 57
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
Mask-Group	0.434 37	0.778 25	0.516 39	0.471 19	0.330 34	0.658 34	0.029 48	0.526 29	0.249 34	0.256 41	0.400 35	0.309 33	0.384 48	0.296 65	0.368 27	0.575 39	0.425 36	0.877 33	0.362 24
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022
Occipital-SCS	0.320 57	0.679 43	0.352 58	0.334 48	0.229 51	0.436 59	0.025 49	0.412 45	0.058 60	0.161 63	0.240 56	0.085 59	0.262 54	0.496 54	0.187 52	0.467 57	0.328 55	0.775 52	0.231 51

SALoss-ResNet	0.262 60	0.667 44	0.335 59	0.067 69	0.123 63	0.427 61	0.022 50	0.280 64	0.058 59	0.216 50	0.211 60	0.039 65	0.142 65	0.519 51	0.106 66	0.338 66	0.310 58	0.721 60	0.138 61
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)
SSEN	0.384 47	0.852 17	0.494 47	0.192 59	0.226 52	0.648 38	0.022 51	0.398 47	0.299 29	0.277 37	0.317 47	0.231 45	0.194 62	0.514 52	0.196 49	0.586 37	0.444 32	0.843 46	0.184 55
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv
PE	0.396 43	0.667 44	0.467 49	0.446 25	0.243 49	0.624 46	0.022 52	0.577 15	0.106 52	0.219 48	0.340 43	0.239 42	0.487 27	0.475 56	0.225 46	0.541 48	0.350 49	0.818 50	0.273 44
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
PointGroup	0.407 41	0.639 51	0.496 45	0.415 33	0.243 50	0.645 39	0.021 53	0.570 17	0.114 51	0.211 53	0.359 42	0.217 47	0.428 38	0.660 40	0.256 42	0.562 44	0.341 51	0.860 38	0.291 37
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]
RWSeg	0.348 54	0.475 61	0.456 52	0.320 50	0.275 44	0.476 57	0.020 54	0.491 35	0.056 62	0.212 52	0.320 46	0.261 39	0.302 53	0.520 50	0.182 53	0.557 45	0.285 61	0.867 36	0.197 53

Mask3D_evaluation	0.382 48	0.593 54	0.520 37	0.390 38	0.314 37	0.600 47	0.018 55	0.287 63	0.151 44	0.281 35	0.387 38	0.169 53	0.429 36	0.654 41	0.172 55	0.578 38	0.384 43	0.670 62	0.278 42

RPGN	0.428 39	0.630 52	0.508 43	0.367 43	0.249 48	0.658 35	0.016 56	0.673 4	0.131 48	0.234 45	0.383 40	0.270 38	0.434 35	0.748 22	0.274 39	0.609 30	0.406 38	0.842 47	0.267 46
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
R-PointNet	0.158 68	0.356 67	0.173 71	0.113 64	0.140 61	0.359 64	0.012 57	0.023 74	0.039 63	0.134 64	0.123 66	0.008 72	0.089 68	0.149 70	0.117 64	0.221 71	0.128 71	0.563 67	0.094 65

NeuralBF	0.353 53	0.593 54	0.511 42	0.375 41	0.264 45	0.597 48	0.008 58	0.332 54	0.160 43	0.229 47	0.274 53	0.000 76	0.206 59	0.678 36	0.155 59	0.485 55	0.422 37	0.816 51	0.254 48
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
Sgpn_scannet	0.049 76	0.023 77	0.134 75	0.031 74	0.013 76	0.144 76	0.006 59	0.008 76	0.000 76	0.028 76	0.017 75	0.003 73	0.009 77	0.000 76	0.021 73	0.122 76	0.095 76	0.175 74	0.054 74

PCJC	0.375 49	0.704 37	0.542 33	0.284 53	0.197 56	0.649 37	0.006 60	0.426 42	0.138 46	0.242 43	0.304 49	0.183 52	0.388 47	0.629 42	0.141 62	0.546 47	0.344 50	0.738 57	0.283 41

3D-BEVIS	0.117 71	0.250 71	0.308 64	0.020 75	0.009 77	0.269 73	0.006 61	0.008 75	0.029 65	0.037 75	0.014 76	0.003 74	0.036 73	0.147 71	0.042 70	0.381 63	0.118 72	0.362 71	0.069 70
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
Sem_Recon_ins	0.098 74	0.295 70	0.187 70	0.015 76	0.036 73	0.213 75	0.005 62	0.038 73	0.003 72	0.056 71	0.037 71	0.036 67	0.015 75	0.051 72	0.044 69	0.209 74	0.098 75	0.354 72	0.071 69

AOIA	0.387 46	0.704 37	0.515 40	0.385 39	0.225 53	0.669 31	0.005 63	0.482 38	0.126 49	0.181 58	0.269 54	0.221 46	0.426 41	0.478 55	0.218 47	0.592 34	0.371 44	0.851 39	0.242 49

UNet-backbone	0.161 66	0.519 59	0.259 66	0.084 65	0.059 68	0.325 70	0.002 64	0.093 71	0.009 70	0.077 69	0.064 68	0.045 64	0.044 72	0.161 69	0.045 68	0.331 67	0.180 66	0.566 66	0.033 76

3D-SIS	0.161 66	0.407 65	0.155 73	0.068 68	0.043 72	0.346 67	0.001 65	0.134 68	0.005 71	0.088 66	0.106 67	0.037 66	0.135 67	0.321 64	0.028 72	0.339 65	0.116 73	0.466 69	0.093 66
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
MTML	0.282 59	0.577 57	0.380 54	0.182 61	0.107 65	0.430 60	0.001 66	0.422 43	0.057 61	0.179 59	0.162 62	0.070 62	0.229 57	0.511 53	0.161 56	0.491 54	0.313 56	0.650 65	0.162 58
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
MaskRCNN 2d->3d Proj	0.022 77	0.185 74	0.000 77	0.000 77	0.015 75	0.000 77	0.000 67	0.006 77	0.000 76	0.010 77	0.006 77	0.107 58	0.012 76	0.000 76	0.002 77	0.027 77	0.004 77	0.022 77	0.001 77

MASC	0.254 61	0.463 63	0.249 67	0.113 63	0.167 58	0.412 63	0.000 68	0.374 51	0.073 56	0.173 61	0.243 55	0.130 56	0.228 58	0.368 61	0.160 57	0.356 64	0.208 64	0.711 61	0.136 62
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
PanopticFusion-inst	0.214 65	0.250 71	0.330 61	0.275 54	0.103 66	0.228 74	0.000 69	0.345 52	0.024 66	0.088 67	0.203 61	0.186 51	0.167 64	0.367 62	0.125 63	0.221 72	0.112 74	0.666 63	0.162 59
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 70	0.296 69	0.203 69	0.071 67	0.058 69	0.349 66	0.000 69	0.150 67	0.019 67	0.054 72	0.034 73	0.017 71	0.052 70	0.042 74	0.013 75	0.209 73	0.183 65	0.371 70	0.057 72

Region-18class	0.146 69	0.175 75	0.321 63	0.080 66	0.062 67	0.357 65	0.000 69	0.307 58	0.002 73	0.066 70	0.044 70	0.000 76	0.018 74	0.036 75	0.054 67	0.447 61	0.133 69	0.472 68	0.060 71

tmp	0.113 73	0.333 68	0.151 74	0.056 72	0.053 70	0.344 69	0.000 69	0.105 70	0.016 69	0.049 73	0.035 72	0.020 70	0.053 69	0.048 73	0.013 74	0.183 75	0.173 67	0.344 73	0.054 73

SegGroup_ins	0.246 64	0.556 58	0.335 60	0.062 71	0.115 64	0.490 56	0.000 69	0.297 61	0.018 68	0.186 57	0.142 63	0.083 60	0.233 56	0.216 67	0.153 60	0.469 56	0.251 62	0.744 56	0.083 67
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
One_Thing_One_Click	0.326 56	0.472 62	0.361 57	0.232 57	0.183 57	0.555 53	0.000 69	0.498 34	0.038 64	0.195 56	0.226 58	0.362 27	0.168 63	0.469 57	0.251 43	0.553 46	0.335 53	0.846 44	0.117 64
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
SPG_WSIS	0.251 63	0.380 66	0.274 65	0.289 52	0.144 60	0.413 62	0.000 69	0.311 57	0.065 58	0.113 65	0.130 65	0.029 68	0.204 60	0.388 60	0.108 65	0.459 58	0.311 57	0.769 53	0.127 63

Hier3D	0.117 71	0.222 73	0.161 72	0.054 73	0.027 74	0.289 71	0.000 69	0.124 69	0.001 75	0.079 68	0.061 69	0.027 69	0.141 66	0.240 66	0.005 76	0.310 68	0.129 70	0.153 76	0.081 68
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
ASIS	0.085 75	0.037 76	0.080 76	0.066 70	0.047 71	0.282 72	0.000 69	0.052 72	0.002 74	0.047 74	0.026 74	0.001 75	0.046 71	0.194 68	0.031 71	0.264 70	0.140 68	0.167 75	0.047 75

3D Semantic Instance Benchmark