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

SIM3D	0.617 3	0.952 4	0.629 16	0.539 11	0.426 15	0.768 10	0.302 6	0.681 2	0.425 9	0.473 15	0.511 15	0.701 2	0.717 1	0.821 6	0.467 14	0.774 1	0.559 14	0.914 17	0.448 2

OneFormer3D	0.566 16	0.781 23	0.697 6	0.562 5	0.431 14	0.770 8	0.331 3	0.400 45	0.373 17	0.529 2	0.504 19	0.568 17	0.475 27	0.732 25	0.470 12	0.762 2	0.550 18	0.871 34	0.379 16
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation.
GraphCut	0.552 19	1.000 1	0.611 18	0.438 27	0.392 21	0.714 22	0.139 23	0.598 13	0.327 21	0.389 22	0.510 16	0.598 11	0.427 38	0.754 20	0.463 17	0.761 3	0.588 8	0.903 22	0.329 28

PointRel	0.622 1	0.926 8	0.710 3	0.541 10	0.502 2	0.772 6	0.314 4	0.598 11	0.425 8	0.504 9	0.565 1	0.650 6	0.716 2	0.809 7	0.476 11	0.747 4	0.618 1	0.963 3	0.364 19
: Relation3D (PointRel): Enhancing Relation Modeling for Point Cloud Instance Segmentation.
EV3D	0.615 5	0.946 5	0.652 13	0.555 6	0.433 13	0.773 5	0.271 12	0.604 8	0.447 5	0.506 6	0.544 6	0.698 3	0.716 2	0.775 15	0.480 8	0.747 5	0.572 12	0.925 13	0.435 5

Spherical Mask(CtoF)	0.616 4	0.946 5	0.654 12	0.555 6	0.434 12	0.769 9	0.271 11	0.604 8	0.447 5	0.505 7	0.549 2	0.698 3	0.716 2	0.775 15	0.480 8	0.747 5	0.575 10	0.925 13	0.436 4

Queryformer	0.583 11	0.926 8	0.702 5	0.393 36	0.504 1	0.733 20	0.276 10	0.527 28	0.373 16	0.479 14	0.534 10	0.533 22	0.697 7	0.720 28	0.436 20	0.745 7	0.592 6	0.958 6	0.363 20

TST3D	0.569 15	0.778 24	0.675 9	0.598 3	0.451 10	0.727 21	0.280 9	0.476 38	0.395 11	0.472 16	0.457 27	0.583 13	0.580 15	0.777 12	0.462 18	0.735 8	0.547 19	0.919 16	0.333 27
Duc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee: MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation. ACM Multimedia 2024
ExtMask3D	0.598 6	0.852 16	0.692 7	0.433 30	0.461 7	0.791 3	0.264 13	0.488 35	0.493 2	0.508 5	0.528 14	0.594 12	0.706 6	0.791 9	0.483 6	0.734 9	0.595 4	0.911 19	0.437 3

PBNet	0.573 14	0.926 8	0.575 25	0.619 2	0.472 5	0.736 18	0.239 17	0.487 36	0.383 14	0.459 18	0.506 18	0.533 21	0.585 13	0.767 17	0.404 23	0.717 10	0.559 15	0.969 2	0.381 15
W.Zhao, Y.Yan, C.Yang, J.Ye,X.Yang,K.Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023
InsSSM	0.586 10	1.000 1	0.593 20	0.440 26	0.480 4	0.771 7	0.345 1	0.437 39	0.444 7	0.495 13	0.548 4	0.579 15	0.621 10	0.720 27	0.409 22	0.712 11	0.593 5	0.960 4	0.395 9
Lei Yao, Yi Wang, Moyun Liu, Lap-Pui Chau: SGIFormer: Semantic-guided and Geometric-enhanced Interleaving Transformer for 3D Instance Segmentation. TCSVT, 2024
SPFormer	0.549 20	0.745 27	0.640 14	0.484 16	0.395 19	0.739 16	0.311 5	0.566 18	0.335 20	0.468 17	0.492 20	0.555 19	0.478 26	0.747 22	0.436 19	0.712 12	0.540 20	0.893 26	0.343 26
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral]
Box2Mask	0.433 36	0.741 29	0.463 49	0.433 29	0.283 39	0.625 43	0.103 26	0.298 58	0.125 48	0.260 38	0.424 32	0.322 29	0.472 28	0.701 33	0.363 26	0.711 13	0.309 57	0.882 29	0.272 43
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022
SoftGroup++	0.513 23	0.704 35	0.578 24	0.398 35	0.363 29	0.704 23	0.061 35	0.647 5	0.297 28	0.378 25	0.537 9	0.343 26	0.614 12	0.828 5	0.295 33	0.710 14	0.505 25	0.875 33	0.394 10

Competitor-SPFormer	0.580 13	0.721 33	0.705 4	0.593 4	0.444 11	0.786 4	0.286 8	0.564 19	0.376 15	0.498 11	0.534 11	0.546 20	0.390 43	0.785 11	0.577 1	0.708 15	0.579 9	0.954 7	0.388 12

Competitor-MAFT	0.618 2	0.866 15	0.724 1	0.628 1	0.484 3	0.803 1	0.300 7	0.509 32	0.496 1	0.539 1	0.547 5	0.703 1	0.668 8	0.708 30	0.463 16	0.708 16	0.595 3	0.959 5	0.418 7

DANCENET	0.504 25	0.926 8	0.579 21	0.472 18	0.367 26	0.626 42	0.165 22	0.432 40	0.221 33	0.408 21	0.449 29	0.411 24	0.564 16	0.746 23	0.421 21	0.707 17	0.438 32	0.846 42	0.288 38

UniPerception	0.588 8	0.963 3	0.667 10	0.493 15	0.472 6	0.750 14	0.229 18	0.528 27	0.468 4	0.498 12	0.542 7	0.643 7	0.530 20	0.661 37	0.463 15	0.695 18	0.599 2	0.972 1	0.420 6

MAFT	0.596 7	0.889 13	0.721 2	0.448 23	0.460 8	0.768 11	0.251 15	0.558 21	0.408 10	0.504 8	0.539 8	0.616 10	0.618 11	0.858 3	0.482 7	0.684 19	0.551 17	0.931 12	0.450 1

TD3D	0.489 27	0.852 16	0.511 39	0.434 28	0.322 33	0.735 19	0.101 28	0.512 30	0.355 19	0.349 26	0.468 25	0.283 34	0.514 23	0.676 36	0.268 38	0.671 20	0.510 22	0.908 20	0.329 29
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
DualGroup	0.469 30	0.815 20	0.552 28	0.398 34	0.374 24	0.683 28	0.130 24	0.539 24	0.310 22	0.327 29	0.407 33	0.276 35	0.447 31	0.535 47	0.342 27	0.659 21	0.455 30	0.900 25	0.301 33

ISBNet	0.559 18	0.939 7	0.655 11	0.383 39	0.426 16	0.763 12	0.180 20	0.534 25	0.386 12	0.499 10	0.509 17	0.621 9	0.427 37	0.704 32	0.467 13	0.649 22	0.571 13	0.948 9	0.401 8
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
KmaxOneFormerNet	0.581 12	0.745 27	0.692 8	0.551 8	0.458 9	0.798 2	0.264 14	0.531 26	0.369 18	0.513 4	0.531 13	0.632 8	0.494 24	0.798 8	0.567 2	0.648 23	0.558 16	0.950 8	0.362 21

Mask3D	0.566 16	0.926 8	0.597 19	0.408 33	0.420 17	0.737 17	0.239 16	0.598 11	0.386 13	0.458 19	0.549 2	0.568 18	0.716 2	0.601 43	0.480 8	0.646 24	0.575 10	0.922 15	0.364 18
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
SoftGroup	0.504 25	0.667 42	0.579 22	0.372 41	0.381 22	0.694 25	0.072 31	0.677 3	0.303 24	0.387 23	0.531 12	0.319 30	0.582 14	0.754 19	0.318 29	0.643 25	0.492 26	0.907 21	0.388 13
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral]
DKNet	0.532 21	0.815 20	0.624 17	0.517 12	0.377 23	0.749 15	0.107 25	0.509 31	0.304 23	0.437 20	0.475 22	0.581 14	0.539 18	0.775 14	0.339 28	0.640 26	0.506 23	0.901 23	0.385 14
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022
SSTNet	0.506 24	0.738 31	0.549 30	0.497 14	0.316 34	0.693 26	0.178 21	0.377 48	0.198 37	0.330 27	0.463 26	0.576 16	0.515 22	0.857 4	0.494 3	0.637 27	0.457 29	0.943 10	0.290 37
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021
TopoSeg	0.479 29	0.704 35	0.564 26	0.467 20	0.366 27	0.633 40	0.068 32	0.554 22	0.262 30	0.328 28	0.447 30	0.323 28	0.534 19	0.722 26	0.288 35	0.614 28	0.482 27	0.912 18	0.358 24

RPGN	0.428 37	0.630 50	0.508 41	0.367 42	0.249 46	0.658 34	0.016 54	0.673 4	0.131 46	0.234 43	0.383 38	0.270 36	0.434 33	0.748 21	0.274 37	0.609 29	0.406 37	0.842 46	0.267 44
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022
MG-Former	0.587 9	0.852 16	0.639 15	0.454 22	0.393 20	0.758 13	0.338 2	0.572 16	0.480 3	0.527 3	0.491 21	0.671 5	0.527 21	0.867 1	0.485 5	0.601 30	0.590 7	0.938 11	0.390 11

OccuSeg+instance	0.486 28	0.802 22	0.536 32	0.428 31	0.369 25	0.702 24	0.205 19	0.331 53	0.301 25	0.379 24	0.474 23	0.327 27	0.437 32	0.862 2	0.485 4	0.601 31	0.394 40	0.846 44	0.273 41
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020
AOIA	0.387 44	0.704 35	0.515 38	0.385 38	0.225 51	0.669 30	0.005 61	0.482 37	0.126 47	0.181 56	0.269 52	0.221 44	0.426 39	0.478 53	0.218 45	0.592 32	0.371 43	0.851 38	0.242 47

IPCA-Inst	0.520 22	0.889 13	0.551 29	0.548 9	0.418 18	0.665 32	0.064 34	0.585 14	0.260 31	0.277 36	0.471 24	0.500 23	0.644 9	0.785 10	0.369 24	0.591 33	0.511 21	0.878 31	0.362 22

Dyco3D	0.395 42	0.642 48	0.518 36	0.447 24	0.259 45	0.666 31	0.050 40	0.251 63	0.166 40	0.231 44	0.362 39	0.232 42	0.331 49	0.535 46	0.229 43	0.587 34	0.438 33	0.850 39	0.317 30
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
SSEN	0.384 45	0.852 16	0.494 45	0.192 57	0.226 50	0.648 37	0.022 49	0.398 46	0.299 27	0.277 35	0.317 45	0.231 43	0.194 60	0.514 50	0.196 47	0.586 35	0.444 31	0.843 45	0.184 53
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv
Mask3D_evaluation	0.382 46	0.593 52	0.520 35	0.390 37	0.314 35	0.600 46	0.018 53	0.287 61	0.151 42	0.281 33	0.387 36	0.169 51	0.429 34	0.654 39	0.172 53	0.578 36	0.384 42	0.670 60	0.278 40

Mask-Group	0.434 35	0.778 24	0.516 37	0.471 19	0.330 32	0.658 33	0.029 46	0.526 29	0.249 32	0.256 39	0.400 34	0.309 31	0.384 46	0.296 63	0.368 25	0.575 37	0.425 35	0.877 32	0.362 23
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022
DD-UNet+Group	0.436 33	0.630 50	0.508 42	0.480 17	0.310 36	0.624 44	0.065 33	0.638 6	0.174 39	0.256 40	0.384 37	0.194 47	0.428 35	0.759 18	0.289 34	0.574 38	0.400 38	0.849 41	0.291 36
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
3D-MPA	0.355 50	0.457 62	0.484 46	0.299 49	0.277 41	0.591 49	0.047 42	0.332 51	0.212 35	0.217 47	0.278 48	0.193 48	0.413 42	0.410 57	0.195 48	0.574 39	0.352 46	0.849 40	0.213 50
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020
OSIS	0.392 43	0.778 24	0.530 33	0.220 56	0.278 40	0.567 50	0.083 30	0.330 54	0.299 26	0.270 37	0.310 46	0.143 53	0.260 53	0.624 41	0.277 36	0.568 40	0.361 45	0.865 36	0.301 32

HAIS	0.457 32	0.704 35	0.561 27	0.457 21	0.364 28	0.673 29	0.046 43	0.547 23	0.194 38	0.308 31	0.426 31	0.288 33	0.454 30	0.711 29	0.262 39	0.563 41	0.434 34	0.889 28	0.344 25
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021
PointGroup	0.407 39	0.639 49	0.496 43	0.415 32	0.243 48	0.645 38	0.021 51	0.570 17	0.114 49	0.211 51	0.359 40	0.217 45	0.428 36	0.660 38	0.256 40	0.562 42	0.341 49	0.860 37	0.291 35
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 52	0.475 59	0.456 50	0.320 48	0.275 42	0.476 55	0.020 52	0.491 34	0.056 60	0.212 50	0.320 44	0.261 37	0.302 51	0.520 48	0.182 51	0.557 43	0.285 59	0.867 35	0.197 51

One_Thing_One_Click	0.326 54	0.472 60	0.361 55	0.232 55	0.183 55	0.555 51	0.000 67	0.498 33	0.038 62	0.195 54	0.226 56	0.362 25	0.168 61	0.469 55	0.251 41	0.553 44	0.335 51	0.846 43	0.117 62
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021
PCJC	0.375 47	0.704 35	0.542 31	0.284 51	0.197 54	0.649 36	0.006 58	0.426 41	0.138 44	0.242 41	0.304 47	0.183 50	0.388 45	0.629 40	0.141 60	0.546 45	0.344 48	0.738 55	0.283 39

PE	0.396 41	0.667 42	0.467 47	0.446 25	0.243 47	0.624 45	0.022 50	0.577 15	0.106 50	0.219 46	0.340 41	0.239 40	0.487 25	0.475 54	0.225 44	0.541 46	0.350 47	0.818 49	0.273 42
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021
INS-Conv-instance	0.435 34	0.716 34	0.495 44	0.355 43	0.331 31	0.689 27	0.102 27	0.394 47	0.208 36	0.280 34	0.395 35	0.250 38	0.544 17	0.741 24	0.309 31	0.536 47	0.391 41	0.842 47	0.258 45

GICN	0.341 53	0.580 54	0.371 54	0.344 45	0.198 53	0.469 56	0.052 38	0.564 20	0.093 51	0.212 49	0.212 57	0.127 55	0.347 48	0.537 45	0.206 46	0.525 48	0.329 52	0.729 56	0.241 48

CSC-Pretrained	0.405 40	0.738 31	0.465 48	0.331 47	0.205 52	0.655 35	0.051 39	0.601 10	0.092 53	0.211 52	0.329 43	0.198 46	0.459 29	0.775 13	0.195 49	0.524 49	0.400 39	0.878 30	0.184 54

ClickSeg_Instance	0.366 48	0.654 46	0.375 53	0.184 58	0.302 37	0.592 48	0.050 41	0.300 57	0.093 52	0.283 32	0.277 49	0.249 39	0.426 40	0.615 42	0.299 32	0.504 50	0.367 44	0.832 48	0.191 52

SSEC	0.465 31	0.667 42	0.578 23	0.502 13	0.362 30	0.641 39	0.035 44	0.605 7	0.291 29	0.323 30	0.451 28	0.296 32	0.417 41	0.677 35	0.245 42	0.501 51	0.506 24	0.900 24	0.366 17

MTML	0.282 57	0.577 55	0.380 52	0.182 59	0.107 63	0.430 58	0.001 64	0.422 42	0.057 59	0.179 57	0.162 60	0.070 60	0.229 55	0.511 51	0.161 54	0.491 52	0.313 54	0.650 63	0.162 56
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral]
NeuralBF	0.353 51	0.593 52	0.511 40	0.375 40	0.264 43	0.597 47	0.008 56	0.332 52	0.160 41	0.229 45	0.274 51	0.000 74	0.206 57	0.678 34	0.155 57	0.485 53	0.422 36	0.816 50	0.254 46
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
SegGroup_ins	0.246 62	0.556 56	0.335 58	0.062 69	0.115 62	0.490 54	0.000 67	0.297 59	0.018 66	0.186 55	0.142 61	0.083 58	0.233 54	0.216 65	0.153 58	0.469 54	0.251 60	0.744 54	0.083 65
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022
Occipital-SCS	0.320 55	0.679 41	0.352 56	0.334 46	0.229 49	0.436 57	0.025 47	0.412 44	0.058 58	0.161 61	0.240 54	0.085 57	0.262 52	0.496 52	0.187 50	0.467 55	0.328 53	0.775 51	0.231 49

SPG_WSIS	0.251 61	0.380 64	0.274 63	0.289 50	0.144 58	0.413 60	0.000 67	0.311 55	0.065 56	0.113 63	0.130 63	0.029 66	0.204 58	0.388 58	0.108 63	0.459 56	0.311 55	0.769 52	0.127 61

SphereSeg	0.357 49	0.651 47	0.411 51	0.345 44	0.264 44	0.630 41	0.059 36	0.289 60	0.212 34	0.240 42	0.336 42	0.158 52	0.305 50	0.557 44	0.159 56	0.455 57	0.341 50	0.726 57	0.294 34

DENet	0.413 38	0.741 29	0.520 34	0.237 54	0.284 38	0.523 53	0.097 29	0.691 1	0.138 43	0.209 53	0.229 55	0.238 41	0.390 44	0.707 31	0.310 30	0.448 58	0.470 28	0.892 27	0.310 31

Region-18class	0.146 67	0.175 73	0.321 61	0.080 64	0.062 65	0.357 63	0.000 67	0.307 56	0.002 71	0.066 68	0.044 68	0.000 74	0.018 72	0.036 73	0.054 65	0.447 59	0.133 67	0.472 66	0.060 69

Sparse R-CNN	0.292 56	0.704 35	0.213 66	0.153 60	0.154 57	0.551 52	0.053 37	0.212 64	0.132 45	0.174 58	0.274 50	0.070 59	0.363 47	0.441 56	0.176 52	0.424 60	0.234 61	0.758 53	0.161 58

3D-BEVIS	0.117 69	0.250 69	0.308 62	0.020 73	0.009 75	0.269 71	0.006 59	0.008 73	0.029 63	0.037 73	0.014 74	0.003 72	0.036 71	0.147 69	0.042 68	0.381 61	0.118 70	0.362 69	0.069 68
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation.
MASC	0.254 59	0.463 61	0.249 65	0.113 61	0.167 56	0.412 61	0.000 66	0.374 49	0.073 54	0.173 59	0.243 53	0.130 54	0.228 56	0.368 59	0.160 55	0.356 62	0.208 62	0.711 59	0.136 60
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation.
3D-SIS	0.161 64	0.407 63	0.155 71	0.068 66	0.043 70	0.346 65	0.001 63	0.134 66	0.005 69	0.088 64	0.106 65	0.037 64	0.135 65	0.321 62	0.028 70	0.339 63	0.116 71	0.466 67	0.093 64
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019
SALoss-ResNet	0.262 58	0.667 42	0.335 57	0.067 67	0.123 61	0.427 59	0.022 48	0.280 62	0.058 57	0.216 48	0.211 58	0.039 63	0.142 63	0.519 49	0.106 64	0.338 64	0.310 56	0.721 58	0.138 59
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)
UNet-backbone	0.161 64	0.519 57	0.259 64	0.084 63	0.059 66	0.325 68	0.002 62	0.093 69	0.009 68	0.077 67	0.064 66	0.045 62	0.044 70	0.161 67	0.045 66	0.331 65	0.180 64	0.566 64	0.033 74

Hier3D	0.117 69	0.222 71	0.161 70	0.054 71	0.027 72	0.289 69	0.000 67	0.124 67	0.001 73	0.079 66	0.061 67	0.027 67	0.141 64	0.240 64	0.005 74	0.310 66	0.129 68	0.153 74	0.081 66
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation.
3D-BoNet	0.253 60	0.519 57	0.324 60	0.251 53	0.137 60	0.345 66	0.031 45	0.419 43	0.069 55	0.162 60	0.131 62	0.052 61	0.202 59	0.338 61	0.147 59	0.301 67	0.303 58	0.651 62	0.178 55
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
ASIS	0.085 73	0.037 74	0.080 74	0.066 68	0.047 69	0.282 70	0.000 67	0.052 70	0.002 72	0.047 72	0.026 72	0.001 73	0.046 69	0.194 66	0.031 69	0.264 68	0.140 66	0.167 73	0.047 73

R-PointNet	0.158 66	0.356 65	0.173 69	0.113 62	0.140 59	0.359 62	0.012 55	0.023 72	0.039 61	0.134 62	0.123 64	0.008 70	0.089 66	0.149 68	0.117 62	0.221 69	0.128 69	0.563 65	0.094 63

PanopticFusion-inst	0.214 63	0.250 69	0.330 59	0.275 52	0.103 64	0.228 72	0.000 67	0.345 50	0.024 64	0.088 65	0.203 59	0.186 49	0.167 62	0.367 60	0.125 61	0.221 70	0.112 72	0.666 61	0.162 57
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 68	0.296 67	0.203 67	0.071 65	0.058 67	0.349 64	0.000 67	0.150 65	0.019 65	0.054 70	0.034 71	0.017 69	0.052 68	0.042 72	0.013 73	0.209 71	0.183 63	0.371 68	0.057 70

Sem_Recon_ins	0.098 72	0.295 68	0.187 68	0.015 74	0.036 71	0.213 73	0.005 60	0.038 71	0.003 70	0.056 69	0.037 69	0.036 65	0.015 73	0.051 70	0.044 67	0.209 72	0.098 73	0.354 70	0.071 67

tmp	0.113 71	0.333 66	0.151 72	0.056 70	0.053 68	0.344 67	0.000 67	0.105 68	0.016 67	0.049 71	0.035 70	0.020 68	0.053 67	0.048 71	0.013 72	0.183 73	0.173 65	0.344 71	0.054 71

Sgpn_scannet	0.049 74	0.023 75	0.134 73	0.031 72	0.013 74	0.144 74	0.006 57	0.008 74	0.000 74	0.028 74	0.017 73	0.003 71	0.009 75	0.000 74	0.021 71	0.122 74	0.095 74	0.175 72	0.054 72

MaskRCNN 2d->3d Proj	0.022 75	0.185 72	0.000 75	0.000 75	0.015 73	0.000 75	0.000 65	0.006 75	0.000 74	0.010 75	0.006 75	0.107 56	0.012 74	0.000 74	0.002 75	0.027 75	0.004 75	0.022 75	0.001 75

3D Semantic Instance Benchmark