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 50% | bathtub | bed | bookshelf | cabinet | chair | counter | curtain | desk | door | otherfurniture | picture | refrigerator | shower curtain | sink | sofa | table | toilet | window |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
MAFT | 0.786 6 | 1.000 1 | 0.894 11 | 0.807 16 | 0.694 10 | 0.893 7 | 0.486 7 | 0.674 25 | 0.740 7 | 0.786 1 | 0.704 7 | 0.727 4 | 0.739 10 | 1.000 1 | 0.707 16 | 0.849 9 | 0.756 16 | 1.000 1 | 0.685 1 | |
OneFormer3D | 0.801 2 | 1.000 1 | 0.973 2 | 0.909 4 | 0.698 8 | 0.928 2 | 0.582 1 | 0.668 27 | 0.685 11 | 0.780 2 | 0.687 8 | 0.698 10 | 0.702 11 | 1.000 1 | 0.794 6 | 0.900 2 | 0.784 9 | 0.986 44 | 0.635 4 | |
Maxim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: OneFormer3D: One Transformer for Unified Point Cloud Segmentation. | ||||||||||||||||||||
SIM3D | 0.766 10 | 1.000 1 | 0.948 4 | 0.582 47 | 0.599 25 | 0.882 11 | 0.510 5 | 0.701 21 | 0.632 18 | 0.772 3 | 0.685 9 | 0.687 12 | 0.782 7 | 1.000 1 | 0.833 2 | 0.756 31 | 0.798 5 | 1.000 1 | 0.622 7 | |
ExtMask3D | 0.789 4 | 1.000 1 | 0.988 1 | 0.756 26 | 0.706 6 | 0.912 4 | 0.429 12 | 0.647 32 | 0.806 4 | 0.755 4 | 0.673 10 | 0.689 11 | 0.772 9 | 1.000 1 | 0.789 7 | 0.852 7 | 0.811 3 | 1.000 1 | 0.617 9 | |
UniPerception | 0.800 3 | 1.000 1 | 0.930 6 | 0.872 8 | 0.727 2 | 0.862 16 | 0.454 11 | 0.764 13 | 0.820 1 | 0.746 5 | 0.706 6 | 0.750 2 | 0.772 8 | 0.926 37 | 0.764 10 | 0.818 20 | 0.826 1 | 0.997 34 | 0.660 2 | |
Spherical Mask(CtoF) | 0.812 1 | 1.000 1 | 0.973 3 | 0.852 10 | 0.718 3 | 0.917 3 | 0.574 2 | 0.677 24 | 0.748 6 | 0.729 6 | 0.715 4 | 0.795 1 | 0.809 1 | 1.000 1 | 0.831 3 | 0.854 6 | 0.787 7 | 1.000 1 | 0.638 3 | |
SPFormer | 0.770 8 | 0.903 48 | 0.903 8 | 0.806 17 | 0.609 23 | 0.886 8 | 0.568 3 | 0.815 6 | 0.705 10 | 0.711 7 | 0.655 11 | 0.652 16 | 0.685 16 | 1.000 1 | 0.789 8 | 0.809 21 | 0.776 12 | 1.000 1 | 0.583 18 | |
Sun Jiahao, Qing Chunmei, Tan Junpeng, Xu Xiangmin: Superpoint Transformer for 3D Scene Instance Segmentation. AAAI 2023 [Oral] | ||||||||||||||||||||
ISBNet | 0.757 12 | 1.000 1 | 0.904 7 | 0.731 29 | 0.678 12 | 0.895 5 | 0.458 9 | 0.644 34 | 0.670 13 | 0.710 8 | 0.620 18 | 0.732 3 | 0.650 18 | 1.000 1 | 0.756 11 | 0.778 24 | 0.779 10 | 1.000 1 | 0.614 10 | |
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 | ||||||||||||||||||||
DKNet | 0.718 18 | 1.000 1 | 0.814 21 | 0.782 20 | 0.619 20 | 0.872 12 | 0.224 25 | 0.751 16 | 0.569 24 | 0.677 9 | 0.585 24 | 0.724 5 | 0.633 29 | 0.981 31 | 0.515 39 | 0.819 18 | 0.736 19 | 1.000 1 | 0.617 8 | |
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong: 3D Instances as 1D Kernels. ECCV 2022 | ||||||||||||||||||||
Mask3D | 0.780 7 | 1.000 1 | 0.786 35 | 0.716 31 | 0.696 9 | 0.885 10 | 0.500 6 | 0.714 19 | 0.810 3 | 0.672 10 | 0.715 4 | 0.679 13 | 0.809 1 | 1.000 1 | 0.831 3 | 0.833 13 | 0.787 7 | 1.000 1 | 0.602 13 | |
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023 | ||||||||||||||||||||
Queryformer | 0.787 5 | 1.000 1 | 0.933 5 | 0.601 41 | 0.754 1 | 0.886 9 | 0.558 4 | 0.661 29 | 0.767 5 | 0.665 11 | 0.716 3 | 0.639 17 | 0.808 3 | 1.000 1 | 0.844 1 | 0.897 3 | 0.804 4 | 1.000 1 | 0.624 6 | |
SoftGroup | 0.761 11 | 1.000 1 | 0.808 24 | 0.845 11 | 0.716 4 | 0.862 15 | 0.243 24 | 0.824 4 | 0.655 16 | 0.620 12 | 0.734 2 | 0.699 9 | 0.791 6 | 0.981 31 | 0.716 14 | 0.844 10 | 0.769 13 | 1.000 1 | 0.594 16 | |
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo: SoftGroup for 3D Instance Segmentaiton on Point Clouds. CVPR 2022 [Oral] | ||||||||||||||||||||
PBNet | 0.747 14 | 1.000 1 | 0.818 20 | 0.837 13 | 0.713 5 | 0.844 18 | 0.457 10 | 0.647 32 | 0.711 9 | 0.614 13 | 0.617 20 | 0.657 15 | 0.650 18 | 1.000 1 | 0.692 17 | 0.822 16 | 0.765 15 | 1.000 1 | 0.595 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 | ||||||||||||||||||||
DANCENET | 0.680 23 | 1.000 1 | 0.807 25 | 0.733 28 | 0.600 24 | 0.768 37 | 0.375 16 | 0.543 44 | 0.538 25 | 0.610 14 | 0.599 21 | 0.498 27 | 0.632 31 | 0.981 31 | 0.739 13 | 0.856 5 | 0.633 32 | 0.882 55 | 0.454 39 | |
TD3D | 0.751 13 | 1.000 1 | 0.774 36 | 0.867 9 | 0.621 19 | 0.934 1 | 0.404 13 | 0.706 20 | 0.812 2 | 0.605 15 | 0.633 16 | 0.626 18 | 0.690 15 | 1.000 1 | 0.640 25 | 0.820 17 | 0.777 11 | 1.000 1 | 0.612 11 | |
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024 | ||||||||||||||||||||
GraphCut | 0.732 15 | 1.000 1 | 0.788 33 | 0.724 30 | 0.642 17 | 0.859 17 | 0.248 23 | 0.787 11 | 0.618 20 | 0.596 16 | 0.653 13 | 0.722 6 | 0.583 39 | 1.000 1 | 0.766 9 | 0.861 4 | 0.825 2 | 1.000 1 | 0.504 30 | |
SoftGroup++ | 0.769 9 | 1.000 1 | 0.803 28 | 0.937 1 | 0.684 11 | 0.865 13 | 0.213 27 | 0.870 2 | 0.664 14 | 0.571 17 | 0.758 1 | 0.702 8 | 0.807 4 | 1.000 1 | 0.653 23 | 0.902 1 | 0.792 6 | 1.000 1 | 0.626 5 | |
OccuSeg+instance | 0.672 26 | 1.000 1 | 0.758 44 | 0.682 34 | 0.576 29 | 0.842 19 | 0.477 8 | 0.504 50 | 0.524 26 | 0.567 18 | 0.585 26 | 0.451 31 | 0.557 41 | 1.000 1 | 0.751 12 | 0.797 22 | 0.563 42 | 1.000 1 | 0.467 38 | |
Lei Han, Tian Zheng, Lan Xu, Lu Fang: OccuSeg: Occupancy-aware 3D Instance Segmentation. CVPR2020 | ||||||||||||||||||||
Box2Mask | 0.677 25 | 1.000 1 | 0.847 15 | 0.771 22 | 0.509 40 | 0.816 24 | 0.277 20 | 0.558 43 | 0.482 27 | 0.562 19 | 0.640 14 | 0.448 32 | 0.700 12 | 1.000 1 | 0.666 18 | 0.852 8 | 0.578 39 | 0.997 34 | 0.488 34 | |
Julian Chibane, Francis Engelmann, Tuan Anh Tran, Gerard Pons-Moll: Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV 2022 | ||||||||||||||||||||
SSTNet | 0.698 21 | 1.000 1 | 0.697 52 | 0.888 6 | 0.556 31 | 0.803 29 | 0.387 14 | 0.626 36 | 0.417 35 | 0.556 20 | 0.585 25 | 0.702 7 | 0.600 33 | 1.000 1 | 0.824 5 | 0.720 40 | 0.692 21 | 1.000 1 | 0.509 29 | |
Zhihao Liang, Zhihao Li, Songcen Xu, Mingkui Tan, Kui Jia: Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks. ICCV2021 | ||||||||||||||||||||
TopoSeg | 0.725 17 | 1.000 1 | 0.806 27 | 0.933 2 | 0.668 14 | 0.758 38 | 0.272 21 | 0.734 18 | 0.630 19 | 0.549 21 | 0.654 12 | 0.606 19 | 0.697 14 | 0.966 34 | 0.612 29 | 0.839 11 | 0.754 17 | 1.000 1 | 0.573 19 | |
SphereSeg | 0.680 23 | 1.000 1 | 0.856 12 | 0.744 27 | 0.618 21 | 0.893 6 | 0.151 30 | 0.651 31 | 0.713 8 | 0.537 22 | 0.579 27 | 0.430 36 | 0.651 17 | 1.000 1 | 0.389 50 | 0.744 35 | 0.697 20 | 0.991 42 | 0.601 14 | |
SSEC | 0.707 19 | 1.000 1 | 0.850 13 | 0.924 3 | 0.648 15 | 0.747 41 | 0.162 29 | 0.862 3 | 0.572 23 | 0.520 23 | 0.624 17 | 0.549 22 | 0.649 27 | 1.000 1 | 0.560 34 | 0.706 41 | 0.768 14 | 1.000 1 | 0.591 17 | |
HAIS | 0.699 20 | 1.000 1 | 0.849 14 | 0.820 14 | 0.675 13 | 0.808 28 | 0.279 19 | 0.757 15 | 0.465 30 | 0.517 24 | 0.596 22 | 0.559 21 | 0.600 33 | 1.000 1 | 0.654 22 | 0.767 27 | 0.676 23 | 0.994 40 | 0.560 23 | |
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang: Hierarchical Aggregation for 3D Instance Segmentation. ICCV 2021 | ||||||||||||||||||||
IPCA-Inst | 0.731 16 | 1.000 1 | 0.788 34 | 0.884 7 | 0.698 7 | 0.788 34 | 0.252 22 | 0.760 14 | 0.646 17 | 0.511 25 | 0.637 15 | 0.665 14 | 0.804 5 | 1.000 1 | 0.644 24 | 0.778 25 | 0.747 18 | 1.000 1 | 0.561 22 | |
Mask3D_evaluation | 0.631 36 | 1.000 1 | 0.829 18 | 0.606 40 | 0.646 16 | 0.836 20 | 0.068 43 | 0.511 48 | 0.462 31 | 0.507 26 | 0.619 19 | 0.389 41 | 0.610 32 | 1.000 1 | 0.432 45 | 0.828 15 | 0.673 24 | 0.788 59 | 0.552 24 | |
OSIS | 0.605 39 | 1.000 1 | 0.801 29 | 0.599 42 | 0.535 34 | 0.728 43 | 0.286 18 | 0.436 54 | 0.679 12 | 0.491 27 | 0.433 42 | 0.256 47 | 0.404 54 | 0.857 39 | 0.620 28 | 0.724 38 | 0.510 52 | 1.000 1 | 0.539 25 | |
PCJC | 0.578 41 | 1.000 1 | 0.810 23 | 0.583 46 | 0.449 47 | 0.813 27 | 0.042 48 | 0.603 39 | 0.341 43 | 0.490 28 | 0.465 39 | 0.410 38 | 0.650 18 | 0.835 47 | 0.264 56 | 0.694 45 | 0.561 43 | 0.889 52 | 0.504 31 | |
DualGroup | 0.694 22 | 1.000 1 | 0.799 30 | 0.811 15 | 0.622 18 | 0.817 23 | 0.376 15 | 0.805 9 | 0.590 22 | 0.487 29 | 0.568 28 | 0.525 26 | 0.650 18 | 0.835 47 | 0.600 30 | 0.829 14 | 0.655 26 | 1.000 1 | 0.526 26 | |
DD-UNet+Group | 0.635 35 | 0.667 50 | 0.797 32 | 0.714 32 | 0.562 30 | 0.774 36 | 0.146 31 | 0.810 8 | 0.429 34 | 0.476 30 | 0.546 33 | 0.399 39 | 0.633 29 | 1.000 1 | 0.632 26 | 0.722 39 | 0.609 33 | 1.000 1 | 0.514 27 | |
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 | ||||||||||||||||||||
INS-Conv-instance | 0.657 28 | 1.000 1 | 0.760 42 | 0.667 36 | 0.581 27 | 0.863 14 | 0.323 17 | 0.655 30 | 0.477 28 | 0.473 31 | 0.549 31 | 0.432 35 | 0.650 18 | 1.000 1 | 0.655 21 | 0.738 36 | 0.585 38 | 0.944 47 | 0.472 37 | |
SSEN | 0.575 42 | 1.000 1 | 0.761 40 | 0.473 50 | 0.477 45 | 0.795 32 | 0.066 44 | 0.529 46 | 0.658 15 | 0.460 32 | 0.461 40 | 0.380 42 | 0.331 56 | 0.859 38 | 0.401 49 | 0.692 47 | 0.653 27 | 1.000 1 | 0.348 48 | |
Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim: Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning. Arxiv | ||||||||||||||||||||
Mask-Group | 0.664 27 | 1.000 1 | 0.822 19 | 0.764 25 | 0.616 22 | 0.815 25 | 0.139 34 | 0.694 23 | 0.597 21 | 0.459 33 | 0.566 29 | 0.599 20 | 0.600 33 | 0.516 57 | 0.715 15 | 0.819 19 | 0.635 30 | 1.000 1 | 0.603 12 | |
Min Zhong, Xinghao Chen, Xiaokang Chen, Gang Zeng, Yunhe Wang: MaskGroup: Hierarchical Point Grouping and Masking for 3D Instance Segmentation. ICME 2022 | ||||||||||||||||||||
GICN | 0.638 33 | 1.000 1 | 0.895 10 | 0.800 18 | 0.480 44 | 0.676 46 | 0.144 32 | 0.737 17 | 0.354 41 | 0.447 34 | 0.400 48 | 0.365 43 | 0.700 12 | 1.000 1 | 0.569 32 | 0.836 12 | 0.599 34 | 1.000 1 | 0.473 36 | |
PointGroup | 0.636 34 | 1.000 1 | 0.765 39 | 0.624 38 | 0.505 42 | 0.797 31 | 0.116 38 | 0.696 22 | 0.384 39 | 0.441 35 | 0.559 30 | 0.476 29 | 0.596 36 | 1.000 1 | 0.666 18 | 0.756 32 | 0.556 46 | 0.997 34 | 0.513 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] | ||||||||||||||||||||
NeuralBF | 0.555 44 | 0.667 50 | 0.896 9 | 0.843 12 | 0.517 39 | 0.751 40 | 0.029 49 | 0.519 47 | 0.414 36 | 0.439 36 | 0.465 38 | 0.000 66 | 0.484 45 | 0.857 39 | 0.287 54 | 0.693 46 | 0.651 29 | 1.000 1 | 0.485 35 | |
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 | ||||||||||||||||||||
Dyco3D | 0.641 32 | 1.000 1 | 0.841 16 | 0.893 5 | 0.531 35 | 0.802 30 | 0.115 39 | 0.588 41 | 0.448 32 | 0.438 37 | 0.537 34 | 0.430 37 | 0.550 42 | 0.857 39 | 0.534 37 | 0.764 29 | 0.657 25 | 0.987 43 | 0.568 20 | |
Tong He; Chunhua Shen; Anton van den Hengel: DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021 | ||||||||||||||||||||
3D-MPA | 0.611 38 | 1.000 1 | 0.833 17 | 0.765 24 | 0.526 37 | 0.756 39 | 0.136 36 | 0.588 41 | 0.470 29 | 0.438 38 | 0.432 44 | 0.358 45 | 0.650 18 | 0.857 39 | 0.429 46 | 0.765 28 | 0.557 45 | 1.000 1 | 0.430 41 | |
Francis Engelmann, Martin Bokeloh, Alireza Fathi, Bastian Leibe, Matthias Nießner: 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation. CVPR 2020 | ||||||||||||||||||||
RPGN | 0.643 31 | 1.000 1 | 0.758 43 | 0.582 48 | 0.539 32 | 0.826 22 | 0.046 47 | 0.765 12 | 0.372 40 | 0.436 39 | 0.588 23 | 0.539 25 | 0.650 18 | 1.000 1 | 0.577 31 | 0.750 33 | 0.653 28 | 0.997 34 | 0.495 33 | |
Shichao Dong, Guosheng Lin, Tzu-Yi Hung: Learning Regional Purity for Instance Segmentation on 3D Point Clouds. ECCV 2022 | ||||||||||||||||||||
PE | 0.645 30 | 1.000 1 | 0.773 38 | 0.798 19 | 0.538 33 | 0.786 35 | 0.088 42 | 0.799 10 | 0.350 42 | 0.435 40 | 0.547 32 | 0.545 23 | 0.646 28 | 0.933 36 | 0.562 33 | 0.761 30 | 0.556 47 | 0.997 34 | 0.501 32 | |
Biao Zhang, Peter Wonka: Point Cloud Instance Segmentation using Probabilistic Embeddings. CVPR 2021 | ||||||||||||||||||||
ClickSeg_Instance | 0.539 46 | 1.000 1 | 0.621 55 | 0.300 53 | 0.530 36 | 0.698 44 | 0.127 37 | 0.533 45 | 0.222 49 | 0.430 41 | 0.400 47 | 0.365 43 | 0.574 40 | 0.938 35 | 0.472 42 | 0.659 50 | 0.543 48 | 0.944 47 | 0.347 49 | |
CSC-Pretrained | 0.648 29 | 1.000 1 | 0.810 22 | 0.768 23 | 0.523 38 | 0.813 26 | 0.143 33 | 0.819 5 | 0.389 38 | 0.422 42 | 0.511 35 | 0.443 33 | 0.650 18 | 1.000 1 | 0.624 27 | 0.732 37 | 0.634 31 | 1.000 1 | 0.375 46 | |
MTML | 0.549 45 | 1.000 1 | 0.807 26 | 0.588 45 | 0.327 52 | 0.647 48 | 0.004 56 | 0.815 7 | 0.180 51 | 0.418 43 | 0.364 52 | 0.182 50 | 0.445 48 | 1.000 1 | 0.442 44 | 0.688 48 | 0.571 41 | 1.000 1 | 0.396 44 | |
Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald: 3D Instance Segmentation via Multi-task Metric Learning. ICCV 2019 [oral] | ||||||||||||||||||||
SALoss-ResNet | 0.459 53 | 1.000 1 | 0.737 46 | 0.159 64 | 0.259 56 | 0.587 53 | 0.138 35 | 0.475 52 | 0.217 50 | 0.416 44 | 0.408 46 | 0.128 53 | 0.315 57 | 0.714 50 | 0.411 47 | 0.536 59 | 0.590 37 | 0.873 56 | 0.304 51 | |
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) | ||||||||||||||||||||
Sparse R-CNN | 0.515 48 | 1.000 1 | 0.538 60 | 0.282 54 | 0.468 46 | 0.790 33 | 0.173 28 | 0.345 56 | 0.429 33 | 0.413 45 | 0.484 37 | 0.176 51 | 0.595 38 | 0.591 55 | 0.522 38 | 0.668 49 | 0.476 53 | 0.986 45 | 0.327 50 | |
AOIA | 0.601 40 | 1.000 1 | 0.761 41 | 0.687 33 | 0.485 43 | 0.828 21 | 0.008 54 | 0.663 28 | 0.405 37 | 0.405 46 | 0.425 45 | 0.490 28 | 0.596 36 | 0.714 50 | 0.553 36 | 0.779 23 | 0.597 35 | 0.992 41 | 0.424 43 | |
DENet | 0.629 37 | 1.000 1 | 0.797 31 | 0.608 39 | 0.589 26 | 0.627 50 | 0.219 26 | 0.882 1 | 0.310 44 | 0.402 47 | 0.383 50 | 0.396 40 | 0.650 18 | 1.000 1 | 0.663 20 | 0.543 58 | 0.691 22 | 1.000 1 | 0.568 21 | |
RWSeg | 0.567 43 | 0.528 60 | 0.708 51 | 0.626 37 | 0.580 28 | 0.745 42 | 0.063 45 | 0.627 35 | 0.240 48 | 0.400 48 | 0.497 36 | 0.464 30 | 0.515 43 | 1.000 1 | 0.475 41 | 0.745 34 | 0.571 40 | 1.000 1 | 0.429 42 | |
One_Thing_One_Click | 0.529 47 | 0.667 50 | 0.718 47 | 0.777 21 | 0.399 48 | 0.683 45 | 0.000 59 | 0.669 26 | 0.138 54 | 0.391 49 | 0.374 51 | 0.539 24 | 0.360 55 | 0.641 54 | 0.556 35 | 0.774 26 | 0.593 36 | 0.997 34 | 0.251 54 | |
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 | 0.445 55 | 0.667 50 | 0.773 37 | 0.185 61 | 0.317 53 | 0.656 47 | 0.000 59 | 0.407 55 | 0.134 55 | 0.381 50 | 0.267 55 | 0.217 49 | 0.476 46 | 0.714 50 | 0.452 43 | 0.629 53 | 0.514 50 | 1.000 1 | 0.222 57 | |
An Tao, Yueqi Duan, Yi Wei, Jiwen Lu, Jie Zhou: SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation. TIP 2022 | ||||||||||||||||||||
MASC | 0.447 54 | 0.528 60 | 0.555 58 | 0.381 51 | 0.382 49 | 0.633 49 | 0.002 57 | 0.509 49 | 0.260 46 | 0.361 51 | 0.432 43 | 0.327 46 | 0.451 47 | 0.571 56 | 0.367 52 | 0.639 52 | 0.386 54 | 0.980 46 | 0.276 53 | |
Chen Liu, Yasutaka Furukawa: MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation. | ||||||||||||||||||||
Occipital-SCS | 0.512 49 | 1.000 1 | 0.716 48 | 0.509 49 | 0.506 41 | 0.611 51 | 0.092 41 | 0.602 40 | 0.177 52 | 0.346 52 | 0.383 49 | 0.165 52 | 0.442 49 | 0.850 46 | 0.386 51 | 0.618 54 | 0.543 49 | 0.889 52 | 0.389 45 | |
3D-BoNet | 0.488 50 | 1.000 1 | 0.672 54 | 0.590 44 | 0.301 54 | 0.484 61 | 0.098 40 | 0.620 37 | 0.306 45 | 0.341 53 | 0.259 56 | 0.125 54 | 0.434 51 | 0.796 49 | 0.402 48 | 0.499 60 | 0.513 51 | 0.909 51 | 0.439 40 | |
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 | ||||||||||||||||||||
3D-SIS | 0.382 56 | 1.000 1 | 0.432 63 | 0.245 56 | 0.190 58 | 0.577 54 | 0.013 53 | 0.263 58 | 0.033 61 | 0.320 54 | 0.240 57 | 0.075 57 | 0.422 52 | 0.857 39 | 0.117 61 | 0.699 43 | 0.271 62 | 0.883 54 | 0.235 56 | |
Ji Hou, Angela Dai, Matthias Niessner: 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans. CVPR 2019 | ||||||||||||||||||||
SPG_WSIS | 0.470 52 | 0.667 50 | 0.685 53 | 0.677 35 | 0.372 50 | 0.562 55 | 0.000 59 | 0.482 51 | 0.244 47 | 0.316 55 | 0.298 53 | 0.052 61 | 0.442 50 | 0.857 39 | 0.267 55 | 0.702 42 | 0.559 44 | 1.000 1 | 0.287 52 | |
R-PointNet | 0.306 59 | 0.500 62 | 0.405 64 | 0.311 52 | 0.348 51 | 0.589 52 | 0.054 46 | 0.068 64 | 0.126 56 | 0.283 56 | 0.290 54 | 0.028 62 | 0.219 60 | 0.214 61 | 0.331 53 | 0.396 64 | 0.275 60 | 0.821 58 | 0.245 55 | |
Hier3D | 0.323 57 | 0.667 50 | 0.542 59 | 0.264 55 | 0.157 61 | 0.550 56 | 0.000 59 | 0.205 61 | 0.009 63 | 0.270 57 | 0.218 58 | 0.075 57 | 0.500 44 | 0.688 53 | 0.007 67 | 0.698 44 | 0.301 59 | 0.459 64 | 0.200 58 | |
Tan: HCFS3D: Hierarchical Coupled Feature Selection Network for 3D Semantic and Instance Segmentation. | ||||||||||||||||||||
PanopticFusion-inst | 0.478 51 | 0.667 50 | 0.712 50 | 0.595 43 | 0.259 57 | 0.550 57 | 0.000 59 | 0.613 38 | 0.175 53 | 0.250 58 | 0.434 41 | 0.437 34 | 0.411 53 | 0.857 39 | 0.485 40 | 0.591 57 | 0.267 63 | 0.944 47 | 0.359 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) | ||||||||||||||||||||
Region-18class | 0.284 60 | 0.250 66 | 0.751 45 | 0.228 59 | 0.270 55 | 0.521 58 | 0.000 59 | 0.468 53 | 0.008 65 | 0.205 59 | 0.127 60 | 0.000 66 | 0.068 64 | 0.070 65 | 0.262 57 | 0.652 51 | 0.323 57 | 0.740 60 | 0.173 59 | |
UNet-backbone | 0.319 58 | 0.667 50 | 0.715 49 | 0.233 57 | 0.189 59 | 0.479 62 | 0.008 54 | 0.218 59 | 0.067 60 | 0.201 60 | 0.173 59 | 0.107 55 | 0.123 62 | 0.438 58 | 0.150 58 | 0.615 55 | 0.355 55 | 0.916 50 | 0.093 66 | |
Sem_Recon_ins | 0.227 64 | 0.764 49 | 0.486 61 | 0.069 66 | 0.098 64 | 0.426 64 | 0.017 52 | 0.067 65 | 0.015 62 | 0.172 61 | 0.100 61 | 0.096 56 | 0.054 66 | 0.183 62 | 0.135 59 | 0.366 65 | 0.260 64 | 0.614 63 | 0.168 61 | |
tmp | 0.248 62 | 0.667 50 | 0.437 62 | 0.188 60 | 0.153 62 | 0.491 60 | 0.000 59 | 0.208 60 | 0.094 59 | 0.153 62 | 0.099 63 | 0.057 59 | 0.217 61 | 0.119 63 | 0.039 63 | 0.466 61 | 0.302 58 | 0.640 62 | 0.140 63 | |
SemRegionNet-20cls | 0.250 61 | 0.333 63 | 0.613 56 | 0.229 58 | 0.163 60 | 0.493 59 | 0.000 59 | 0.304 57 | 0.107 57 | 0.147 63 | 0.100 62 | 0.052 60 | 0.231 58 | 0.119 63 | 0.039 63 | 0.445 62 | 0.325 56 | 0.654 61 | 0.141 62 | |
ASIS | 0.199 65 | 0.333 63 | 0.253 66 | 0.167 63 | 0.140 63 | 0.438 63 | 0.000 59 | 0.177 62 | 0.008 64 | 0.121 64 | 0.069 64 | 0.004 65 | 0.231 59 | 0.429 59 | 0.036 65 | 0.445 63 | 0.273 61 | 0.333 66 | 0.119 65 | |
3D-BEVIS | 0.248 62 | 0.667 50 | 0.566 57 | 0.076 65 | 0.035 67 | 0.394 65 | 0.027 51 | 0.035 66 | 0.098 58 | 0.099 65 | 0.030 66 | 0.025 63 | 0.098 63 | 0.375 60 | 0.126 60 | 0.604 56 | 0.181 65 | 0.854 57 | 0.171 60 | |
Cathrin Elich, Francis Engelmann, Jonas Schult, Theodora Kontogianni, Bastian Leibe: 3D-BEVIS: Birds-Eye-View Instance Segmentation. | ||||||||||||||||||||
Sgpn_scannet | 0.143 66 | 0.208 67 | 0.390 65 | 0.169 62 | 0.065 65 | 0.275 66 | 0.029 50 | 0.069 63 | 0.000 66 | 0.087 66 | 0.043 65 | 0.014 64 | 0.027 67 | 0.000 66 | 0.112 62 | 0.351 66 | 0.168 66 | 0.438 65 | 0.138 64 | |
MaskRCNN 2d->3d Proj | 0.058 67 | 0.333 63 | 0.002 67 | 0.000 67 | 0.053 66 | 0.002 67 | 0.002 58 | 0.021 67 | 0.000 66 | 0.045 67 | 0.024 67 | 0.238 48 | 0.065 65 | 0.000 66 | 0.014 66 | 0.107 67 | 0.020 67 | 0.110 67 | 0.006 67 | |