The 3D semantic instance prediction task involves detecting and segmenting the object in an 3D scan mesh.

Evaluation and metrics

Similarly to the ScanNet benchmark in ScanNet200 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) for all 200 categories. Note that multiple predictions of the same ground truth instance are penalized as false positives.



This table lists the benchmark results for the ScanNet200 3D semantic instance scenario.




Method Infoavg ap 25%head ap 25%common ap 25%tail ap 25%alarm clockarmchairbackpackbagballbarbasketbathroom cabinetbathroom counterbathroom stallbathroom stall doorbathroom vanitybathtubbedbenchbicyclebinblackboardblanketblindsboardbookbookshelfbottlebowlboxbroombucketbulletin boardcabinetcalendarcandlecartcase of water bottlescd caseceilingceiling lightchairclockclosetcloset doorcloset rodcloset wallclothesclothes dryercoat rackcoffee kettlecoffee makercoffee tablecolumncomputer towercontainercopiercouchcountercratecupcurtaincushiondecorationdeskdining tabledish rackdishwasherdividerdoordoorframedresserdumbbelldustpanend tablefanfile cabinetfire alarmfire extinguisherfireplacefolded chairfurnitureguitarguitar casehair dryerhandicap barhatheadphonesironing boardjacketkeyboardkeyboard pianokitchen cabinetkitchen counterladderlamplaptoplaundry basketlaundry detergentlaundry hamperledgelightlight switchluggagemachinemailboxmatmattressmicrowavemini fridgemirrormonitormousemusic standnightstandobjectoffice chairottomanovenpaperpaper bagpaper cutterpaper towel dispenserpaper towel rollpersonpianopicturepillarpillowpipeplantplateplungerposterpotted plantpower outletpower stripprinterprojectorprojector screenpurserackradiatorrailrange hoodrecycling binrefrigeratorscaleseatshelfshoeshowershower curtainshower curtain rodshower doorshower floorshower headshower wallsignsinksoap dishsoap dispensersofa chairspeakerstair railstairsstandstoolstorage binstorage containerstorage organizerstovestructurestuffed animalsuitcasetabletelephonetissue boxtoastertoaster oventoilettoilet papertoilet paper dispensertoilet paper holdertoilet seat cover dispensertoweltrash bintrash cantraytubetvtv standvacuum cleanerventwardrobewashing machinewater bottlewater coolerwater pitcherwhiteboardwindowwindowsill
sort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysorted bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort by
AQ3D-ScanNet2000.527 30.743 10.477 30.324 40.862 20.161 30.500 10.500 10.833 10.000 10.513 60.556 70.873 11.000 10.955 70.567 41.000 10.025 50.533 60.435 10.333 50.458 30.556 10.754 40.279 60.167 50.396 30.000 10.083 10.607 30.631 10.250 70.000 10.500 10.000 40.901 30.073 70.930 10.009 80.113 70.803 30.360 20.667 40.056 30.000 10.489 90.805 50.250 70.938 11.000 10.947 30.087 80.368 30.858 10.160 30.026 80.893 20.023 70.750 10.600 10.886 40.843 20.778 20.458 70.089 80.479 50.991 10.143 30.318 60.200 50.000 20.621 50.500 10.366 40.895 41.000 10.839 40.837 20.738 40.707 30.434 60.000 20.000 10.517 30.105 50.465 10.000 50.563 21.000 10.400 30.831 60.658 40.750 40.960 50.000 50.000 10.896 50.088 80.328 21.000 10.071 20.667 11.000 10.868 20.083 60.428 51.000 10.925 20.795 20.337 20.879 20.000 70.714 20.000 20.141 80.000 40.667 60.000 31.000 10.281 10.838 51.000 10.534 20.929 31.000 10.030 20.410 40.602 41.000 11.000 11.000 10.714 81.000 10.712 80.000 30.926 50.819 60.392 30.172 30.707 31.000 10.000 50.576 30.016 50.000 40.000 20.778 20.000 20.000 10.671 40.784 40.768 31.000 10.736 21.000 10.549 61.000 10.835 30.434 30.935 20.000 10.083 20.764 91.000 11.000 10.000 10.664 11.000 10.000 10.333 90.667 30.546 70.816 30.923 1
Mask3D Scannet2000.445 70.653 60.392 70.254 70.648 60.097 50.125 110.000 40.000 60.000 10.657 10.971 10.451 61.000 11.000 10.640 20.500 50.045 31.000 10.241 70.409 40.363 40.440 60.686 80.300 50.000 60.201 70.000 10.009 30.290 70.556 51.000 10.000 10.063 90.000 40.830 60.573 20.844 50.333 30.204 40.058 110.158 110.552 80.056 30.000 11.000 10.725 80.750 10.927 21.000 10.888 80.042 90.120 80.615 100.226 10.250 50.890 30.792 20.677 60.510 60.818 70.699 70.512 80.167 110.125 40.315 80.943 20.309 10.017 90.200 50.000 20.188 60.000 60.183 90.815 71.000 10.827 50.741 60.442 70.414 100.600 20.000 20.000 10.458 40.049 70.321 50.381 10.000 50.908 50.400 30.841 50.260 70.710 50.966 40.265 30.000 10.924 20.152 30.025 50.500 40.027 40.028 61.000 10.556 110.016 70.080 110.500 30.694 80.608 70.084 70.604 80.194 50.538 80.000 20.500 10.000 40.354 100.000 31.000 10.000 50.761 80.930 50.053 100.890 71.000 10.008 30.262 70.358 81.000 11.000 10.792 90.966 21.000 10.765 60.004 20.930 40.780 80.330 50.027 60.625 50.974 50.050 10.412 110.021 40.000 40.000 20.778 20.000 20.000 10.493 70.746 60.454 70.335 70.396 50.930 110.551 51.000 10.552 70.606 10.853 50.000 10.004 60.806 41.000 10.727 70.000 10.042 80.745 70.000 10.399 80.391 50.630 50.721 60.619 4
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
ACGP-ScanNet2000.544 10.737 20.483 20.381 10.747 40.184 20.500 10.500 10.708 20.000 10.371 100.604 60.718 41.000 10.996 40.511 61.000 10.035 41.000 10.388 40.528 30.307 60.517 40.834 10.405 41.000 10.396 40.000 10.042 20.540 50.536 61.000 10.000 10.500 10.050 30.918 20.189 50.801 90.086 50.140 60.667 40.292 41.000 10.000 60.000 10.903 20.838 10.444 60.902 31.000 10.912 50.213 40.367 40.858 20.000 50.396 20.846 71.000 10.677 60.443 70.921 20.905 10.573 60.500 50.064 90.658 30.931 30.144 20.637 20.317 40.000 20.638 40.000 60.462 10.897 31.000 10.860 20.944 10.759 30.682 40.510 30.000 20.000 10.391 60.140 30.462 20.143 40.250 30.884 60.600 20.851 40.677 30.680 70.993 10.000 50.000 10.825 80.185 10.036 41.000 10.089 10.472 21.000 10.765 60.500 30.671 30.500 30.929 10.699 50.349 10.738 50.204 40.782 10.008 10.146 70.024 30.835 30.000 30.250 70.006 40.854 30.974 20.590 10.968 21.000 10.007 40.483 10.666 21.000 11.000 11.000 10.959 31.000 10.867 30.000 30.955 30.850 30.403 20.341 20.637 40.721 80.000 50.552 40.000 80.000 40.000 20.778 20.000 20.000 10.759 10.859 10.767 41.000 10.944 11.000 10.424 70.944 30.840 20.206 80.915 30.000 10.064 30.800 61.000 11.000 10.000 10.461 30.903 30.000 10.507 30.903 10.718 20.842 20.809 2
Rongkun Yang, Ye Zhang, Longguang Wang, Zhiheng Fu, Lian Xu, Yulan Guo: Beyond Context Bias: Adaptive Instance Placement for Robust 3D Instance Segmentation.
DINO3D-Scannet200copyleft0.511 40.685 40.484 10.331 30.864 10.220 10.500 10.000 40.042 50.000 10.576 30.746 30.744 31.000 11.000 10.355 111.000 10.048 20.000 80.327 50.000 60.494 20.532 30.596 100.496 20.250 30.481 10.000 10.000 40.714 10.629 21.000 10.000 10.250 60.663 10.861 50.436 30.892 30.667 10.244 10.385 80.421 11.000 10.000 60.000 10.764 40.719 100.500 20.889 41.000 10.907 60.111 60.378 20.778 40.000 50.595 10.905 10.708 40.750 10.542 20.890 30.754 60.761 30.798 10.220 20.683 20.817 70.000 50.600 30.200 50.500 10.944 10.125 50.334 50.856 60.792 70.873 10.756 50.777 20.803 10.675 10.000 20.000 10.200 80.298 10.412 30.000 50.000 50.719 100.800 10.923 10.750 10.798 30.960 60.000 50.000 10.856 70.142 40.001 80.417 70.000 60.014 71.000 10.824 40.559 10.700 10.500 30.863 40.816 10.163 60.944 10.764 10.714 20.000 20.250 50.000 41.000 10.063 11.000 10.000 50.789 70.974 20.079 90.851 90.000 80.000 60.468 20.702 10.167 51.000 11.000 10.857 40.000 60.867 40.000 30.968 20.845 40.264 80.419 10.500 80.667 100.000 50.677 10.028 30.194 20.000 20.857 10.000 20.000 10.699 30.821 20.930 10.850 40.346 60.944 90.579 10.866 50.850 10.221 70.911 40.000 10.011 50.806 50.764 110.860 50.000 10.472 20.794 50.000 10.667 10.655 40.655 40.811 50.528 5
Jinyuan Qu, Hongyang Li, Xingyu Chen, Shilong Liu, Yukai Shi, Tianhe Ren, Ruitao Jing and Lei Zhang: SegDINO3D: 3D Instance Segmentation Empowered by Both Image-Level and Object-Level 2D Features. AAAI 2026
Volt-SPFormerpermissive0.527 20.731 30.475 40.342 20.789 30.076 60.500 10.000 40.125 30.000 10.391 90.508 100.753 21.000 10.994 50.400 70.500 50.020 71.000 10.413 30.850 20.547 10.510 50.810 20.433 30.250 30.346 50.000 10.000 40.519 60.594 41.000 10.000 10.331 50.000 40.937 10.638 10.826 60.056 70.214 20.850 20.262 50.667 40.028 50.000 10.817 30.825 20.250 70.880 51.000 10.950 20.279 20.309 50.856 30.000 50.304 30.867 50.000 80.750 10.542 20.942 10.818 40.901 10.458 70.329 10.750 10.855 50.000 50.510 40.200 50.000 20.677 20.500 10.397 20.903 21.000 10.843 30.773 41.000 10.799 20.449 50.250 10.000 10.600 10.027 80.372 40.000 51.000 10.833 70.400 30.878 30.656 50.843 20.973 30.000 50.000 10.921 30.103 50.008 60.500 40.057 30.278 41.000 10.802 50.557 20.700 11.000 10.874 30.767 30.279 40.801 40.047 60.714 20.000 20.500 10.250 10.907 20.000 31.000 10.011 30.875 10.944 40.255 40.923 41.000 10.002 50.321 50.579 51.000 11.000 11.000 11.000 11.000 10.737 70.000 30.926 60.857 20.343 40.000 80.741 10.629 110.025 20.500 70.000 80.000 40.000 20.725 80.000 20.000 10.715 20.803 30.738 51.000 10.500 41.000 10.565 30.884 40.812 40.167 90.937 10.000 10.019 40.923 11.000 11.000 10.000 10.099 71.000 10.000 10.472 40.764 20.614 60.815 40.681 3
Kadir Yilmaz, Adrian Kruse, Tristan Höfer, Daan de Geus, Bastian Leibe: Volume Transformer: Revisiting Vanilla Transformers for 3D Scene Understanding.
CompetitorFormer-2000.469 50.676 50.401 60.296 50.692 50.057 80.500 10.083 30.000 60.000 10.534 40.701 50.410 80.903 100.998 30.878 10.500 50.068 10.250 70.424 21.000 10.244 70.556 20.696 60.270 71.000 10.240 60.000 10.000 40.587 40.380 81.000 10.000 10.500 10.000 40.900 40.257 40.901 20.085 60.207 30.863 10.224 61.000 10.109 20.000 10.724 50.806 40.500 20.869 61.000 10.829 90.247 30.474 10.759 60.021 40.269 40.873 40.125 60.467 100.542 20.885 50.829 30.711 40.285 100.118 50.482 40.770 100.025 40.018 80.400 20.000 20.677 20.500 10.222 70.916 11.000 10.818 60.827 30.342 90.650 50.452 40.000 20.000 10.330 70.173 20.278 70.000 50.083 41.000 10.336 70.748 80.508 60.698 60.989 20.286 20.000 10.933 10.175 20.400 10.663 30.015 50.103 51.000 10.829 30.125 40.293 60.500 30.847 50.711 40.295 30.543 100.385 30.581 70.000 20.500 10.000 40.747 40.050 21.000 10.013 20.850 40.886 60.214 80.918 50.125 70.000 60.320 60.610 30.025 110.933 91.000 10.820 60.250 50.901 20.000 30.980 10.878 10.325 60.160 40.574 70.703 90.009 40.540 50.011 60.000 40.000 20.700 90.056 10.000 10.491 80.729 70.617 60.489 60.565 31.000 10.410 90.750 60.629 60.292 50.839 70.000 10.157 10.839 21.000 10.834 60.000 10.131 60.794 50.000 10.667 10.144 90.664 30.854 10.500 6
CSC-Pretrain Inst.permissive0.275 110.466 110.218 100.110 110.625 80.007 110.500 10.000 40.000 60.000 10.000 110.222 110.377 101.000 10.661 110.400 70.000 100.000 80.000 80.119 110.000 60.000 110.277 100.685 90.067 90.000 60.132 90.000 10.000 40.000 100.367 100.000 80.000 10.000 100.000 40.591 90.055 100.783 100.000 90.014 90.500 60.161 100.278 90.000 60.000 10.667 60.768 60.500 20.866 71.000 10.829 100.000 100.019 110.555 110.000 50.000 90.305 110.000 80.750 10.200 100.783 100.429 90.395 90.677 30.020 110.286 90.584 110.000 50.000 100.115 110.000 20.000 80.000 60.145 110.423 110.500 80.364 110.369 100.571 50.448 90.206 110.000 20.000 10.200 80.106 40.065 110.000 50.000 50.750 80.200 80.774 70.000 110.501 90.841 100.000 50.000 10.692 110.063 100.000 90.000 90.000 60.000 90.500 100.649 80.000 80.084 100.125 100.719 60.413 110.004 100.450 110.000 70.638 50.000 20.000 90.000 40.505 90.000 30.000 80.000 50.727 90.833 70.221 60.779 100.000 80.000 60.168 110.311 110.125 70.571 100.500 100.143 110.000 60.250 100.000 30.869 80.667 100.162 110.000 80.250 101.000 10.000 50.500 70.000 80.000 40.000 20.689 100.000 20.000 10.312 100.383 110.114 90.333 80.000 100.997 70.420 80.613 100.212 110.500 20.819 80.000 10.000 70.768 71.000 10.918 40.000 10.000 90.278 110.000 10.333 90.000 110.353 80.546 110.258 10
Ji Hou, Benjamin Graham, Matthias Nießner, Saining Xie: Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts. CVPR 2021
LGround Inst.permissive0.314 90.529 90.225 90.155 90.578 110.010 90.500 10.000 40.000 60.000 10.515 50.556 70.696 51.000 10.927 90.400 70.083 90.000 81.000 10.252 60.000 60.167 80.350 80.731 50.067 90.000 60.123 100.000 10.000 40.036 90.372 90.000 80.000 10.250 60.000 40.569 100.031 110.810 70.000 90.000 100.630 50.183 80.278 90.000 60.000 10.582 80.589 110.500 20.863 81.000 10.940 40.000 100.144 70.716 90.000 50.000 90.484 90.000 80.500 80.400 80.798 90.500 80.278 100.750 20.093 70.166 100.783 80.000 50.200 70.400 20.000 20.000 80.000 60.219 80.539 90.500 80.578 90.413 90.181 110.457 80.375 70.000 20.000 10.050 110.000 90.077 100.000 50.000 50.500 110.000 110.743 90.250 80.488 100.846 90.000 50.000 10.800 90.069 90.000 90.000 90.000 60.000 91.000 10.607 100.000 80.200 70.500 30.694 70.528 80.063 90.659 70.000 70.594 60.000 20.000 90.000 40.571 80.000 30.000 80.000 50.716 100.647 110.221 60.857 80.000 80.000 60.217 90.346 90.071 100.530 111.000 10.429 90.000 60.286 90.000 30.826 110.706 90.208 100.000 80.250 100.744 70.000 50.500 70.042 20.000 40.000 20.746 70.000 20.000 10.517 60.625 90.085 110.333 80.000 101.000 10.378 100.533 110.376 100.042 110.814 90.000 10.000 70.765 81.000 10.600 90.000 10.000 90.667 80.000 10.472 40.333 60.337 90.605 90.305 8
David Rozenberszki, Or Litany, Angela Dai: Language-Grounded Indoor 3D Semantic Segmentation in the Wild.
ODIN - Ins200permissive0.451 60.637 70.407 50.277 60.583 100.116 40.500 10.000 40.125 30.000 10.599 20.823 20.407 90.667 110.941 80.542 51.000 10.000 81.000 10.162 80.000 60.028 100.357 70.695 70.550 10.000 60.475 20.000 10.000 40.714 10.626 31.000 10.000 10.500 10.125 20.749 70.080 60.742 110.528 20.078 80.500 60.334 30.667 40.333 10.000 10.278 110.723 90.250 70.859 91.000 10.826 110.108 70.221 60.763 50.000 50.250 50.742 80.500 50.750 10.400 80.855 60.769 50.701 50.469 60.203 30.406 70.870 40.000 50.963 10.200 50.000 20.000 80.500 10.370 30.886 51.000 10.782 70.504 80.429 80.494 60.337 80.000 20.000 10.600 10.000 90.215 80.226 20.000 50.944 40.200 80.887 20.750 10.874 10.877 80.438 10.000 10.867 60.089 70.003 70.500 40.000 60.333 31.000 10.742 70.125 40.671 30.417 90.616 100.637 60.238 50.873 30.528 20.494 100.000 20.250 50.000 40.688 50.000 31.000 10.000 50.872 20.833 70.275 30.779 101.000 10.000 60.441 30.577 60.167 51.000 10.500 100.777 70.000 60.778 50.000 30.910 70.800 70.232 90.019 70.717 20.833 60.000 50.638 20.284 10.000 40.000 20.778 20.000 20.000 10.597 50.699 80.850 20.333 80.250 80.944 90.571 20.677 80.795 50.264 60.852 60.000 10.000 70.824 31.000 10.668 80.000 10.000 90.667 80.000 10.333 90.333 60.760 10.679 80.404 7
TD3D Scannet200permissive0.379 80.603 80.306 80.190 80.635 70.073 70.500 10.000 40.000 60.000 10.495 70.735 40.275 111.000 10.979 60.590 30.000 100.021 60.000 80.146 90.000 60.356 50.173 110.795 30.226 80.000 60.173 80.000 10.000 40.226 80.390 70.000 80.000 10.250 60.000 40.706 80.061 90.885 40.093 40.186 50.259 100.200 70.667 40.000 60.000 10.667 60.825 30.250 70.834 101.000 10.958 10.553 10.111 90.748 70.220 20.051 70.866 60.792 20.390 110.045 110.800 80.302 110.517 70.533 40.113 60.427 60.843 60.000 50.458 50.600 10.000 20.101 70.000 60.259 60.717 80.500 80.615 80.520 70.526 60.457 70.270 100.000 20.000 10.400 50.088 60.294 60.181 30.000 51.000 10.400 30.710 110.103 90.477 110.905 70.061 40.000 10.906 40.102 60.232 30.125 80.000 60.003 80.792 91.000 10.000 80.102 90.125 100.559 110.523 90.075 80.715 60.000 70.424 110.000 20.396 40.250 10.638 70.000 30.000 80.000 50.622 110.833 70.221 50.970 10.250 60.038 10.260 80.415 70.125 71.000 11.000 10.857 40.000 60.908 10.012 10.869 90.836 50.635 10.111 50.625 51.000 10.020 30.510 60.003 70.009 31.000 10.778 20.000 20.000 10.370 90.755 50.288 80.333 80.274 71.000 10.557 40.731 70.456 80.433 40.769 110.000 10.000 70.621 101.000 10.458 100.000 10.196 50.817 40.000 10.472 40.222 80.205 110.689 70.274 9
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
Minkowski 34D Inst.permissive0.280 100.488 100.192 110.124 100.593 90.010 100.500 10.000 40.000 60.000 10.447 80.535 90.445 71.000 10.861 100.400 70.225 80.000 80.000 80.142 100.000 60.074 90.342 90.467 110.067 90.000 60.119 110.000 10.000 40.000 100.337 110.000 80.000 10.000 100.000 40.506 110.070 80.804 80.000 90.000 100.333 90.172 90.150 110.000 60.000 10.479 100.745 70.000 110.830 111.000 10.904 70.167 50.090 100.732 80.000 50.000 90.443 100.000 80.500 80.542 20.772 110.396 100.077 110.385 90.044 100.118 110.777 90.000 50.000 100.200 50.000 20.000 80.000 60.148 100.502 100.500 80.419 100.159 110.281 100.404 110.317 90.000 20.000 10.200 80.000 90.077 90.000 50.000 50.750 80.200 80.715 100.021 100.551 80.828 110.000 50.000 10.743 100.059 110.000 90.000 90.000 60.000 90.125 110.648 90.000 80.191 80.500 30.669 90.502 100.000 110.568 90.000 70.516 90.000 20.000 90.000 40.305 110.000 30.000 80.000 50.825 60.833 70.021 110.918 50.000 80.000 60.191 100.346 100.100 90.981 81.000 10.286 100.000 60.000 110.000 30.868 100.648 110.292 70.000 80.375 91.000 10.000 50.500 70.000 80.333 10.000 20.538 110.000 20.000 10.213 110.518 100.098 100.528 50.250 80.997 70.284 110.677 80.398 90.167 90.790 100.000 10.000 70.618 110.903 100.200 110.000 10.333 40.333 100.000 10.442 70.083 100.213 100.587 100.131 11
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019