3D Semantic Instance Benchmark
The 3D semantic instance prediction task involves detecting and segmenting the object in an 3D scan mesh.
Evaluation and metricsOur 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 | Info | avg ap | bathtub | bed | bookshelf | cabinet | chair | counter | curtain | desk | door | otherfurniture | picture | refrigerator | shower curtain | sink | sofa | table | toilet | window |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
OneFormer3D | 0.566 8 | 0.781 18 | 0.697 3 | 0.562 3 | 0.431 8 | 0.770 2 | 0.331 1 | 0.400 38 | 0.373 11 | 0.529 1 | 0.504 13 | 0.568 11 | 0.475 21 | 0.732 19 | 0.470 8 | 0.762 1 | 0.550 10 | 0.871 27 | 0.379 11 | |
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation. | ||||||||||||||||||||
ExtMask3D | 0.598 2 | 0.852 12 | 0.692 4 | 0.433 23 | 0.461 4 | 0.791 1 | 0.264 6 | 0.488 29 | 0.493 1 | 0.508 2 | 0.528 8 | 0.594 7 | 0.706 3 | 0.791 5 | 0.483 4 | 0.734 6 | 0.595 2 | 0.911 11 | 0.437 2 | |
Spherical Mask(CtoF) | 0.616 1 | 0.946 3 | 0.654 8 | 0.555 4 | 0.434 7 | 0.769 3 | 0.271 5 | 0.604 7 | 0.447 3 | 0.505 3 | 0.549 1 | 0.698 1 | 0.716 1 | 0.775 10 | 0.480 6 | 0.747 3 | 0.575 5 | 0.925 7 | 0.436 3 | |
MAFT | 0.596 3 | 0.889 9 | 0.721 1 | 0.448 17 | 0.460 5 | 0.768 4 | 0.251 7 | 0.558 16 | 0.408 4 | 0.504 4 | 0.539 4 | 0.616 5 | 0.618 6 | 0.858 2 | 0.482 5 | 0.684 13 | 0.551 9 | 0.931 6 | 0.450 1 | |
ISBNet | 0.559 10 | 0.939 4 | 0.655 7 | 0.383 32 | 0.426 9 | 0.763 5 | 0.180 13 | 0.534 20 | 0.386 6 | 0.499 5 | 0.509 11 | 0.621 4 | 0.427 31 | 0.704 25 | 0.467 9 | 0.649 16 | 0.571 7 | 0.948 4 | 0.401 5 | |
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 | ||||||||||||||||||||
UniPerception | 0.588 4 | 0.963 2 | 0.667 6 | 0.493 9 | 0.472 3 | 0.750 7 | 0.229 10 | 0.528 21 | 0.468 2 | 0.498 6 | 0.542 3 | 0.643 2 | 0.530 16 | 0.661 30 | 0.463 10 | 0.695 12 | 0.599 1 | 0.972 1 | 0.420 4 | |
Queryformer | 0.583 5 | 0.926 5 | 0.702 2 | 0.393 29 | 0.504 1 | 0.733 13 | 0.276 4 | 0.527 22 | 0.373 10 | 0.479 7 | 0.534 6 | 0.533 15 | 0.697 4 | 0.720 21 | 0.436 14 | 0.745 4 | 0.592 3 | 0.958 3 | 0.363 14 | |
SIM3D | 0.550 12 | 0.889 9 | 0.447 43 | 0.487 10 | 0.404 12 | 0.761 6 | 0.214 11 | 0.502 26 | 0.377 9 | 0.476 8 | 0.522 9 | 0.641 3 | 0.561 12 | 0.715 22 | 0.492 2 | 0.627 21 | 0.502 18 | 0.894 18 | 0.387 8 | |
TST3D | 0.569 7 | 0.778 19 | 0.675 5 | 0.598 2 | 0.451 6 | 0.727 14 | 0.280 3 | 0.476 32 | 0.395 5 | 0.472 9 | 0.457 20 | 0.583 8 | 0.580 10 | 0.777 7 | 0.462 12 | 0.735 5 | 0.547 11 | 0.919 9 | 0.333 20 | |
SPFormer | 0.549 13 | 0.745 22 | 0.640 9 | 0.484 11 | 0.395 13 | 0.739 9 | 0.311 2 | 0.566 14 | 0.335 13 | 0.468 10 | 0.492 14 | 0.555 13 | 0.478 20 | 0.747 16 | 0.436 13 | 0.712 8 | 0.540 12 | 0.893 19 | 0.343 19 | |
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral] | ||||||||||||||||||||
PBNet | 0.573 6 | 0.926 5 | 0.575 17 | 0.619 1 | 0.472 2 | 0.736 11 | 0.239 9 | 0.487 30 | 0.383 8 | 0.459 11 | 0.506 12 | 0.533 14 | 0.585 8 | 0.767 11 | 0.404 16 | 0.717 7 | 0.559 8 | 0.969 2 | 0.381 10 | |
W.Zhao, Y.Yan, C.Yang, J.Ye,X.Yang,K.Huang: Divide and Conquer: 3D Instance Segmentation With Point-Wise Binarization. ICCV 2023 | ||||||||||||||||||||
Mask3D | 0.566 8 | 0.926 5 | 0.597 12 | 0.408 26 | 0.420 10 | 0.737 10 | 0.239 8 | 0.598 9 | 0.386 7 | 0.458 12 | 0.549 1 | 0.568 12 | 0.716 1 | 0.601 36 | 0.480 6 | 0.646 17 | 0.575 5 | 0.922 8 | 0.364 13 | |
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023 | ||||||||||||||||||||
DKNet | 0.532 14 | 0.815 15 | 0.624 10 | 0.517 6 | 0.377 16 | 0.749 8 | 0.107 18 | 0.509 25 | 0.304 16 | 0.437 13 | 0.475 15 | 0.581 9 | 0.539 14 | 0.775 9 | 0.339 21 | 0.640 19 | 0.506 15 | 0.901 15 | 0.385 9 | |
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022 | ||||||||||||||||||||
DANCENET | 0.504 18 | 0.926 5 | 0.579 13 | 0.472 13 | 0.367 19 | 0.626 35 | 0.165 15 | 0.432 33 | 0.221 26 | 0.408 14 | 0.449 22 | 0.411 17 | 0.564 11 | 0.746 17 | 0.421 15 | 0.707 11 | 0.438 25 | 0.846 35 | 0.288 31 | |
GraphCut | 0.552 11 | 1.000 1 | 0.611 11 | 0.438 20 | 0.392 14 | 0.714 15 | 0.139 16 | 0.598 10 | 0.327 14 | 0.389 15 | 0.510 10 | 0.598 6 | 0.427 32 | 0.754 14 | 0.463 11 | 0.761 2 | 0.588 4 | 0.903 14 | 0.329 21 | |
SoftGroup | 0.504 18 | 0.667 35 | 0.579 14 | 0.372 34 | 0.381 15 | 0.694 18 | 0.072 24 | 0.677 2 | 0.303 17 | 0.387 16 | 0.531 7 | 0.319 23 | 0.582 9 | 0.754 13 | 0.318 22 | 0.643 18 | 0.492 19 | 0.907 13 | 0.388 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] | ||||||||||||||||||||
OccuSeg+instance | 0.486 21 | 0.802 17 | 0.536 24 | 0.428 24 | 0.369 18 | 0.702 17 | 0.205 12 | 0.331 46 | 0.301 18 | 0.379 17 | 0.474 16 | 0.327 20 | 0.437 26 | 0.862 1 | 0.485 3 | 0.601 24 | 0.394 33 | 0.846 37 | 0.273 34 | |
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020 | ||||||||||||||||||||
SoftGroup++ | 0.513 16 | 0.704 28 | 0.578 16 | 0.398 28 | 0.363 22 | 0.704 16 | 0.061 28 | 0.647 4 | 0.297 21 | 0.378 18 | 0.537 5 | 0.343 19 | 0.614 7 | 0.828 4 | 0.295 26 | 0.710 10 | 0.505 17 | 0.875 26 | 0.394 6 | |
TD3D | 0.489 20 | 0.852 12 | 0.511 31 | 0.434 21 | 0.322 26 | 0.735 12 | 0.101 21 | 0.512 24 | 0.355 12 | 0.349 19 | 0.468 18 | 0.283 27 | 0.514 18 | 0.676 29 | 0.268 31 | 0.671 14 | 0.510 14 | 0.908 12 | 0.329 22 | |
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024 | ||||||||||||||||||||
SSTNet | 0.506 17 | 0.738 25 | 0.549 22 | 0.497 8 | 0.316 27 | 0.693 19 | 0.178 14 | 0.377 41 | 0.198 30 | 0.330 20 | 0.463 19 | 0.576 10 | 0.515 17 | 0.857 3 | 0.494 1 | 0.637 20 | 0.457 22 | 0.943 5 | 0.290 30 | |
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021 | ||||||||||||||||||||
TopoSeg | 0.479 22 | 0.704 28 | 0.564 18 | 0.467 15 | 0.366 20 | 0.633 33 | 0.068 25 | 0.554 17 | 0.262 23 | 0.328 21 | 0.447 23 | 0.323 21 | 0.534 15 | 0.722 20 | 0.288 28 | 0.614 22 | 0.482 20 | 0.912 10 | 0.358 17 | |
DualGroup | 0.469 23 | 0.815 15 | 0.552 20 | 0.398 27 | 0.374 17 | 0.683 21 | 0.130 17 | 0.539 19 | 0.310 15 | 0.327 22 | 0.407 26 | 0.276 28 | 0.447 25 | 0.535 40 | 0.342 20 | 0.659 15 | 0.455 23 | 0.900 17 | 0.301 26 | |
SSEC | 0.465 24 | 0.667 35 | 0.578 15 | 0.502 7 | 0.362 23 | 0.641 32 | 0.035 37 | 0.605 6 | 0.291 22 | 0.323 23 | 0.451 21 | 0.296 25 | 0.417 35 | 0.677 28 | 0.245 35 | 0.501 44 | 0.506 16 | 0.900 16 | 0.366 12 | |
HAIS | 0.457 25 | 0.704 28 | 0.561 19 | 0.457 16 | 0.364 21 | 0.673 22 | 0.046 36 | 0.547 18 | 0.194 31 | 0.308 24 | 0.426 24 | 0.288 26 | 0.454 24 | 0.711 23 | 0.262 32 | 0.563 34 | 0.434 27 | 0.889 21 | 0.344 18 | |
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021 | ||||||||||||||||||||
ClickSeg_Instance | 0.366 41 | 0.654 39 | 0.375 46 | 0.184 51 | 0.302 30 | 0.592 41 | 0.050 34 | 0.300 50 | 0.093 45 | 0.283 25 | 0.277 42 | 0.249 32 | 0.426 34 | 0.615 35 | 0.299 25 | 0.504 43 | 0.367 37 | 0.832 41 | 0.191 45 | |
Mask3D_evaluation | 0.382 39 | 0.593 45 | 0.520 27 | 0.390 30 | 0.314 28 | 0.600 39 | 0.018 46 | 0.287 54 | 0.151 35 | 0.281 26 | 0.387 29 | 0.169 44 | 0.429 28 | 0.654 32 | 0.172 46 | 0.578 29 | 0.384 35 | 0.670 53 | 0.278 33 | |
INS-Conv-instance | 0.435 27 | 0.716 27 | 0.495 36 | 0.355 36 | 0.331 24 | 0.689 20 | 0.102 20 | 0.394 40 | 0.208 29 | 0.280 27 | 0.395 28 | 0.250 31 | 0.544 13 | 0.741 18 | 0.309 24 | 0.536 40 | 0.391 34 | 0.842 40 | 0.258 38 | |
SSEN | 0.384 38 | 0.852 12 | 0.494 37 | 0.192 50 | 0.226 43 | 0.648 30 | 0.022 42 | 0.398 39 | 0.299 20 | 0.277 28 | 0.317 38 | 0.231 36 | 0.194 53 | 0.514 43 | 0.196 40 | 0.586 28 | 0.444 24 | 0.843 38 | 0.184 46 | |
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv | ||||||||||||||||||||
IPCA-Inst | 0.520 15 | 0.889 9 | 0.551 21 | 0.548 5 | 0.418 11 | 0.665 25 | 0.064 27 | 0.585 11 | 0.260 24 | 0.277 29 | 0.471 17 | 0.500 16 | 0.644 5 | 0.785 6 | 0.369 17 | 0.591 26 | 0.511 13 | 0.878 24 | 0.362 15 | |
OSIS | 0.392 36 | 0.778 19 | 0.530 25 | 0.220 49 | 0.278 33 | 0.567 43 | 0.083 23 | 0.330 47 | 0.299 19 | 0.270 30 | 0.310 39 | 0.143 46 | 0.260 46 | 0.624 34 | 0.277 29 | 0.568 33 | 0.361 38 | 0.865 29 | 0.301 25 | |
Box2Mask | 0.433 29 | 0.741 23 | 0.463 41 | 0.433 22 | 0.283 32 | 0.625 36 | 0.103 19 | 0.298 51 | 0.125 41 | 0.260 31 | 0.424 25 | 0.322 22 | 0.472 22 | 0.701 26 | 0.363 19 | 0.711 9 | 0.309 50 | 0.882 22 | 0.272 36 | |
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022 | ||||||||||||||||||||
Mask-Group | 0.434 28 | 0.778 19 | 0.516 29 | 0.471 14 | 0.330 25 | 0.658 26 | 0.029 39 | 0.526 23 | 0.249 25 | 0.256 32 | 0.400 27 | 0.309 24 | 0.384 39 | 0.296 56 | 0.368 18 | 0.575 30 | 0.425 28 | 0.877 25 | 0.362 16 | |
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 26 | 0.630 43 | 0.508 34 | 0.480 12 | 0.310 29 | 0.624 37 | 0.065 26 | 0.638 5 | 0.174 32 | 0.256 33 | 0.384 30 | 0.194 40 | 0.428 29 | 0.759 12 | 0.289 27 | 0.574 31 | 0.400 31 | 0.849 34 | 0.291 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 | ||||||||||||||||||||
PCJC | 0.375 40 | 0.704 28 | 0.542 23 | 0.284 44 | 0.197 47 | 0.649 29 | 0.006 51 | 0.426 34 | 0.138 37 | 0.242 34 | 0.304 40 | 0.183 43 | 0.388 38 | 0.629 33 | 0.141 53 | 0.546 38 | 0.344 41 | 0.738 48 | 0.283 32 | |
SphereSeg | 0.357 42 | 0.651 40 | 0.411 44 | 0.345 37 | 0.264 37 | 0.630 34 | 0.059 29 | 0.289 53 | 0.212 27 | 0.240 35 | 0.336 35 | 0.158 45 | 0.305 43 | 0.557 37 | 0.159 49 | 0.455 50 | 0.341 43 | 0.726 50 | 0.294 27 | |
RPGN | 0.428 30 | 0.630 43 | 0.508 33 | 0.367 35 | 0.249 39 | 0.658 27 | 0.016 47 | 0.673 3 | 0.131 39 | 0.234 36 | 0.383 31 | 0.270 29 | 0.434 27 | 0.748 15 | 0.274 30 | 0.609 23 | 0.406 30 | 0.842 39 | 0.267 37 | |
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022 | ||||||||||||||||||||
Dyco3D | 0.395 35 | 0.642 41 | 0.518 28 | 0.447 18 | 0.259 38 | 0.666 24 | 0.050 33 | 0.251 56 | 0.166 33 | 0.231 37 | 0.362 32 | 0.232 35 | 0.331 42 | 0.535 39 | 0.229 36 | 0.587 27 | 0.438 26 | 0.850 32 | 0.317 23 | |
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021 | ||||||||||||||||||||
NeuralBF | 0.353 44 | 0.593 45 | 0.511 32 | 0.375 33 | 0.264 36 | 0.597 40 | 0.008 49 | 0.332 45 | 0.160 34 | 0.229 38 | 0.274 44 | 0.000 67 | 0.206 50 | 0.678 27 | 0.155 50 | 0.485 46 | 0.422 29 | 0.816 43 | 0.254 39 | |
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 | ||||||||||||||||||||
PE | 0.396 34 | 0.667 35 | 0.467 39 | 0.446 19 | 0.243 40 | 0.624 38 | 0.022 43 | 0.577 12 | 0.106 43 | 0.219 39 | 0.340 34 | 0.239 33 | 0.487 19 | 0.475 47 | 0.225 37 | 0.541 39 | 0.350 40 | 0.818 42 | 0.273 35 | |
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021 | ||||||||||||||||||||
3D-MPA | 0.355 43 | 0.457 55 | 0.484 38 | 0.299 42 | 0.277 34 | 0.591 42 | 0.047 35 | 0.332 44 | 0.212 28 | 0.217 40 | 0.278 41 | 0.193 41 | 0.413 36 | 0.410 50 | 0.195 41 | 0.574 32 | 0.352 39 | 0.849 33 | 0.213 43 | |
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020 | ||||||||||||||||||||
SALoss-ResNet | 0.262 51 | 0.667 35 | 0.335 50 | 0.067 60 | 0.123 54 | 0.427 52 | 0.022 41 | 0.280 55 | 0.058 50 | 0.216 41 | 0.211 51 | 0.039 56 | 0.142 56 | 0.519 42 | 0.106 57 | 0.338 57 | 0.310 49 | 0.721 51 | 0.138 52 | |
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) | ||||||||||||||||||||
GICN | 0.341 46 | 0.580 47 | 0.371 47 | 0.344 38 | 0.198 46 | 0.469 49 | 0.052 31 | 0.564 15 | 0.093 44 | 0.212 42 | 0.212 50 | 0.127 48 | 0.347 41 | 0.537 38 | 0.206 39 | 0.525 41 | 0.329 45 | 0.729 49 | 0.241 41 | |
RWSeg | 0.348 45 | 0.475 52 | 0.456 42 | 0.320 41 | 0.275 35 | 0.476 48 | 0.020 45 | 0.491 28 | 0.056 53 | 0.212 43 | 0.320 37 | 0.261 30 | 0.302 44 | 0.520 41 | 0.182 44 | 0.557 36 | 0.285 52 | 0.867 28 | 0.197 44 | |
PointGroup | 0.407 32 | 0.639 42 | 0.496 35 | 0.415 25 | 0.243 41 | 0.645 31 | 0.021 44 | 0.570 13 | 0.114 42 | 0.211 44 | 0.359 33 | 0.217 38 | 0.428 30 | 0.660 31 | 0.256 33 | 0.562 35 | 0.341 42 | 0.860 30 | 0.291 28 | |
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] | ||||||||||||||||||||
CSC-Pretrained | 0.405 33 | 0.738 25 | 0.465 40 | 0.331 40 | 0.205 45 | 0.655 28 | 0.051 32 | 0.601 8 | 0.092 46 | 0.211 45 | 0.329 36 | 0.198 39 | 0.459 23 | 0.775 8 | 0.195 42 | 0.524 42 | 0.400 32 | 0.878 23 | 0.184 47 | |
DENet | 0.413 31 | 0.741 23 | 0.520 26 | 0.237 47 | 0.284 31 | 0.523 46 | 0.097 22 | 0.691 1 | 0.138 36 | 0.209 46 | 0.229 48 | 0.238 34 | 0.390 37 | 0.707 24 | 0.310 23 | 0.448 51 | 0.470 21 | 0.892 20 | 0.310 24 | |
One_Thing_One_Click | 0.326 47 | 0.472 53 | 0.361 48 | 0.232 48 | 0.183 48 | 0.555 44 | 0.000 60 | 0.498 27 | 0.038 55 | 0.195 47 | 0.226 49 | 0.362 18 | 0.168 54 | 0.469 48 | 0.251 34 | 0.553 37 | 0.335 44 | 0.846 36 | 0.117 55 | |
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu: One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation. CVPR 2021 | ||||||||||||||||||||
SegGroup_ins | 0.246 55 | 0.556 49 | 0.335 51 | 0.062 62 | 0.115 55 | 0.490 47 | 0.000 60 | 0.297 52 | 0.018 59 | 0.186 48 | 0.142 54 | 0.083 51 | 0.233 47 | 0.216 58 | 0.153 51 | 0.469 47 | 0.251 53 | 0.744 47 | 0.083 58 | |
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022 | ||||||||||||||||||||
AOIA | 0.387 37 | 0.704 28 | 0.515 30 | 0.385 31 | 0.225 44 | 0.669 23 | 0.005 54 | 0.482 31 | 0.126 40 | 0.181 49 | 0.269 45 | 0.221 37 | 0.426 33 | 0.478 46 | 0.218 38 | 0.592 25 | 0.371 36 | 0.851 31 | 0.242 40 | |
MTML | 0.282 50 | 0.577 48 | 0.380 45 | 0.182 52 | 0.107 56 | 0.430 51 | 0.001 57 | 0.422 35 | 0.057 52 | 0.179 50 | 0.162 53 | 0.070 53 | 0.229 48 | 0.511 44 | 0.161 47 | 0.491 45 | 0.313 47 | 0.650 56 | 0.162 49 | |
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral] | ||||||||||||||||||||
Sparse R-CNN | 0.292 49 | 0.704 28 | 0.213 59 | 0.153 53 | 0.154 50 | 0.551 45 | 0.053 30 | 0.212 57 | 0.132 38 | 0.174 51 | 0.274 43 | 0.070 52 | 0.363 40 | 0.441 49 | 0.176 45 | 0.424 53 | 0.234 54 | 0.758 46 | 0.161 51 | |
MASC | 0.254 52 | 0.463 54 | 0.249 58 | 0.113 54 | 0.167 49 | 0.412 54 | 0.000 59 | 0.374 42 | 0.073 47 | 0.173 52 | 0.243 46 | 0.130 47 | 0.228 49 | 0.368 52 | 0.160 48 | 0.356 55 | 0.208 55 | 0.711 52 | 0.136 53 | |
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation. | ||||||||||||||||||||
3D-BoNet | 0.253 53 | 0.519 50 | 0.324 53 | 0.251 46 | 0.137 53 | 0.345 59 | 0.031 38 | 0.419 36 | 0.069 48 | 0.162 53 | 0.131 55 | 0.052 54 | 0.202 52 | 0.338 54 | 0.147 52 | 0.301 60 | 0.303 51 | 0.651 55 | 0.178 48 | |
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 | 0.320 48 | 0.679 34 | 0.352 49 | 0.334 39 | 0.229 42 | 0.436 50 | 0.025 40 | 0.412 37 | 0.058 51 | 0.161 54 | 0.240 47 | 0.085 50 | 0.262 45 | 0.496 45 | 0.187 43 | 0.467 48 | 0.328 46 | 0.775 44 | 0.231 42 | |
R-PointNet | 0.158 59 | 0.356 58 | 0.173 62 | 0.113 55 | 0.140 52 | 0.359 55 | 0.012 48 | 0.023 65 | 0.039 54 | 0.134 55 | 0.123 57 | 0.008 63 | 0.089 59 | 0.149 61 | 0.117 55 | 0.221 62 | 0.128 62 | 0.563 58 | 0.094 56 | |
SPG_WSIS | 0.251 54 | 0.380 57 | 0.274 56 | 0.289 43 | 0.144 51 | 0.413 53 | 0.000 60 | 0.311 48 | 0.065 49 | 0.113 56 | 0.130 56 | 0.029 59 | 0.204 51 | 0.388 51 | 0.108 56 | 0.459 49 | 0.311 48 | 0.769 45 | 0.127 54 | |
3D-SIS | 0.161 57 | 0.407 56 | 0.155 64 | 0.068 59 | 0.043 63 | 0.346 58 | 0.001 56 | 0.134 59 | 0.005 62 | 0.088 57 | 0.106 58 | 0.037 57 | 0.135 58 | 0.321 55 | 0.028 63 | 0.339 56 | 0.116 64 | 0.466 60 | 0.093 57 | |
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019 | ||||||||||||||||||||
PanopticFusion-inst | 0.214 56 | 0.250 62 | 0.330 52 | 0.275 45 | 0.103 57 | 0.228 65 | 0.000 60 | 0.345 43 | 0.024 57 | 0.088 58 | 0.203 52 | 0.186 42 | 0.167 55 | 0.367 53 | 0.125 54 | 0.221 63 | 0.112 65 | 0.666 54 | 0.162 50 | |
Gaku Narita, Takashi Seno, Tomoya Ishikawa, Yohsuke Kaji: PanopticFusion: Online Volumetric Semantic Mapping at the Level of Stuff and Things. IROS 2019 (to appear) | ||||||||||||||||||||
Hier3D | 0.117 62 | 0.222 64 | 0.161 63 | 0.054 64 | 0.027 65 | 0.289 62 | 0.000 60 | 0.124 60 | 0.001 66 | 0.079 59 | 0.061 60 | 0.027 60 | 0.141 57 | 0.240 57 | 0.005 67 | 0.310 59 | 0.129 61 | 0.153 67 | 0.081 59 | |
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation. | ||||||||||||||||||||
UNet-backbone | 0.161 57 | 0.519 50 | 0.259 57 | 0.084 56 | 0.059 59 | 0.325 61 | 0.002 55 | 0.093 62 | 0.009 61 | 0.077 60 | 0.064 59 | 0.045 55 | 0.044 63 | 0.161 60 | 0.045 59 | 0.331 58 | 0.180 57 | 0.566 57 | 0.033 67 | |
Region-18class | 0.146 60 | 0.175 66 | 0.321 54 | 0.080 57 | 0.062 58 | 0.357 56 | 0.000 60 | 0.307 49 | 0.002 64 | 0.066 61 | 0.044 61 | 0.000 67 | 0.018 65 | 0.036 66 | 0.054 58 | 0.447 52 | 0.133 60 | 0.472 59 | 0.060 62 | |
Sem_Recon_ins | 0.098 65 | 0.295 61 | 0.187 61 | 0.015 67 | 0.036 64 | 0.213 66 | 0.005 53 | 0.038 64 | 0.003 63 | 0.056 62 | 0.037 62 | 0.036 58 | 0.015 66 | 0.051 63 | 0.044 60 | 0.209 65 | 0.098 66 | 0.354 63 | 0.071 60 | |
SemRegionNet-20cls | 0.121 61 | 0.296 60 | 0.203 60 | 0.071 58 | 0.058 60 | 0.349 57 | 0.000 60 | 0.150 58 | 0.019 58 | 0.054 63 | 0.034 64 | 0.017 62 | 0.052 61 | 0.042 65 | 0.013 66 | 0.209 64 | 0.183 56 | 0.371 61 | 0.057 63 | |
tmp | 0.113 64 | 0.333 59 | 0.151 65 | 0.056 63 | 0.053 61 | 0.344 60 | 0.000 60 | 0.105 61 | 0.016 60 | 0.049 64 | 0.035 63 | 0.020 61 | 0.053 60 | 0.048 64 | 0.013 65 | 0.183 66 | 0.173 58 | 0.344 64 | 0.054 64 | |
ASIS | 0.085 66 | 0.037 67 | 0.080 67 | 0.066 61 | 0.047 62 | 0.282 63 | 0.000 60 | 0.052 63 | 0.002 65 | 0.047 65 | 0.026 65 | 0.001 66 | 0.046 62 | 0.194 59 | 0.031 62 | 0.264 61 | 0.140 59 | 0.167 66 | 0.047 66 | |
3D-BEVIS | 0.117 62 | 0.250 62 | 0.308 55 | 0.020 66 | 0.009 68 | 0.269 64 | 0.006 52 | 0.008 66 | 0.029 56 | 0.037 66 | 0.014 67 | 0.003 65 | 0.036 64 | 0.147 62 | 0.042 61 | 0.381 54 | 0.118 63 | 0.362 62 | 0.069 61 | |
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation. | ||||||||||||||||||||
Sgpn_scannet | 0.049 67 | 0.023 68 | 0.134 66 | 0.031 65 | 0.013 67 | 0.144 67 | 0.006 50 | 0.008 67 | 0.000 67 | 0.028 67 | 0.017 66 | 0.003 64 | 0.009 68 | 0.000 67 | 0.021 64 | 0.122 67 | 0.095 67 | 0.175 65 | 0.054 65 | |
MaskRCNN 2d->3d Proj | 0.022 68 | 0.185 65 | 0.000 68 | 0.000 68 | 0.015 66 | 0.000 68 | 0.000 58 | 0.006 68 | 0.000 67 | 0.010 68 | 0.006 68 | 0.107 49 | 0.012 67 | 0.000 67 | 0.002 68 | 0.027 68 | 0.004 68 | 0.022 68 | 0.001 68 | |