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 50%head ap 50%common ap 50%tail ap 50%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 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 bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort by
ACGP-ScanNet2000.494 10.656 10.453 20.345 10.723 30.178 20.500 10.500 10.708 10.000 10.020 80.592 30.695 31.000 10.933 50.511 51.000 10.035 41.000 10.378 30.528 30.307 50.442 10.834 10.260 51.000 10.313 20.000 10.042 10.540 40.484 51.000 10.000 10.396 30.000 30.853 10.189 50.771 60.010 20.023 30.471 10.276 20.817 10.000 40.000 10.903 20.838 10.444 50.817 41.000 10.831 70.212 10.310 30.718 10.000 40.396 20.687 41.000 10.677 50.400 10.864 20.545 20.573 50.500 40.062 70.511 30.931 10.144 20.637 20.317 30.000 20.638 30.000 50.396 10.819 41.000 10.655 10.532 30.714 30.680 30.510 30.000 20.000 10.387 50.071 30.405 10.036 40.250 20.449 40.600 20.836 40.677 30.566 20.963 10.000 40.000 10.743 70.148 10.035 31.000 10.089 10.472 11.000 10.765 40.500 20.671 10.500 20.871 10.633 40.040 60.736 30.204 40.627 20.008 10.146 70.024 30.835 20.000 30.250 60.005 10.842 20.974 10.590 10.893 41.000 10.007 30.344 10.495 41.000 11.000 10.792 60.835 51.000 10.643 10.000 30.823 10.844 40.360 20.341 20.088 80.703 50.000 50.548 30.000 20.000 30.000 20.667 50.000 20.000 10.702 20.746 10.664 31.000 10.304 40.997 40.424 60.750 30.840 10.204 70.914 20.000 10.064 20.800 21.000 11.000 10.000 10.461 10.903 20.000 10.504 30.850 10.618 20.633 20.657 1
Volt-SPFormerpermissive0.475 20.630 20.451 30.314 20.772 20.068 50.500 10.000 30.125 20.000 10.107 40.524 60.742 11.000 10.994 10.400 60.500 40.019 51.000 10.410 10.667 20.500 10.423 30.811 20.412 30.250 30.281 30.000 10.000 30.519 50.541 31.000 10.000 10.331 40.000 30.841 30.638 10.806 50.000 40.014 50.241 40.245 30.333 50.028 30.000 10.817 30.825 20.250 60.799 71.000 10.847 40.129 20.294 40.702 30.000 40.304 30.755 10.000 70.750 10.400 10.923 10.482 30.900 10.208 90.319 10.750 10.823 30.000 40.510 30.200 40.000 20.500 40.500 10.300 40.903 11.000 10.564 50.372 51.000 10.787 20.449 50.250 10.000 10.600 10.026 40.375 20.000 51.000 10.455 30.400 30.878 20.641 40.612 10.894 50.000 40.000 10.800 40.078 50.008 50.500 30.056 20.278 31.000 10.797 30.500 20.585 31.000 10.869 20.735 20.056 40.768 20.043 60.714 10.000 20.500 10.250 10.683 50.000 31.000 10.000 20.853 10.944 30.255 30.923 21.000 10.002 40.224 40.499 30.250 31.000 11.000 10.857 21.000 10.613 20.000 30.818 20.857 20.343 30.000 60.209 30.629 90.025 20.500 60.000 20.000 30.000 20.725 30.000 20.000 10.716 10.666 50.651 41.000 10.500 20.990 70.565 20.750 30.699 30.167 80.930 10.000 10.019 30.784 31.000 11.000 10.000 10.099 51.000 10.000 10.472 40.764 20.546 30.621 30.452 4
Kadir Yilmaz, Adrian Kruse, Tristan Höfer, Daan de Geus, Bastian Leibe: Volume Transformer: Revisiting Vanilla Transformers for 3D Scene Understanding.
CompetitorFormer-2000.415 40.574 40.370 50.274 40.632 40.054 70.500 10.083 20.000 50.000 10.260 20.541 50.410 50.903 70.987 20.885 10.500 40.064 10.250 50.378 21.000 10.243 60.428 20.599 70.218 61.000 10.196 50.000 10.000 30.587 30.318 71.000 10.000 10.500 10.000 30.845 20.232 40.885 10.005 30.015 40.216 50.183 60.399 40.082 10.000 10.724 50.806 40.500 20.869 21.000 10.779 80.095 40.443 10.685 50.021 30.269 40.704 30.083 60.467 80.400 10.846 30.551 10.663 30.261 80.103 50.482 40.758 40.025 30.018 70.400 20.000 20.677 20.500 10.207 60.881 21.000 10.600 40.648 20.144 90.641 40.452 40.000 20.000 10.327 60.142 20.209 60.000 50.083 30.215 70.317 60.748 60.508 50.484 40.957 20.000 40.000 10.833 30.132 20.400 10.663 20.015 40.103 41.000 10.759 50.125 40.286 50.500 20.830 30.651 30.089 20.540 80.380 30.581 30.000 20.500 10.000 40.745 30.050 21.000 10.000 20.622 70.694 60.213 50.870 60.125 60.000 50.205 50.562 20.000 80.933 81.000 10.820 60.250 40.347 70.000 30.731 50.877 10.289 40.160 30.186 40.684 60.008 40.538 40.000 20.000 30.000 20.700 40.056 10.000 10.491 50.584 70.602 50.489 40.565 11.000 10.311 80.750 30.583 50.292 40.832 60.000 10.157 10.780 41.000 10.625 50.000 10.131 40.794 40.000 10.667 10.071 90.545 40.682 10.462 3
DINO3D-Scannet200copyleft0.454 30.587 30.453 10.296 30.851 10.200 10.500 10.000 30.042 40.000 10.378 10.545 40.729 21.000 10.981 30.355 101.000 10.046 20.000 70.248 40.000 50.494 20.381 40.586 90.496 20.250 30.409 10.000 10.000 30.714 10.572 11.000 10.000 10.250 50.050 20.793 40.436 30.871 20.000 40.216 10.284 20.290 10.083 70.000 40.000 10.764 40.716 90.500 20.842 31.000 10.891 20.096 30.361 20.690 40.000 40.595 10.753 20.708 40.750 10.400 10.845 40.475 40.728 20.750 10.214 20.683 20.743 50.000 40.400 50.200 40.500 10.944 10.125 40.327 20.823 30.792 60.602 30.662 10.777 20.803 10.675 10.000 20.000 10.200 70.298 10.324 30.000 50.000 40.000 90.800 10.824 50.750 10.507 30.937 30.000 40.000 10.779 60.116 30.001 70.417 60.000 50.014 61.000 10.816 20.548 10.600 20.500 20.771 40.773 10.117 10.944 10.764 10.571 40.000 20.250 50.000 41.000 10.063 11.000 10.000 20.720 40.974 10.079 80.918 30.000 70.000 50.312 30.616 10.125 41.000 11.000 10.857 20.000 50.594 30.000 30.767 30.845 30.264 60.419 10.177 50.667 70.000 50.677 10.000 20.194 10.000 20.857 10.000 20.000 10.563 40.703 20.835 20.850 30.346 30.944 80.499 50.866 20.777 20.221 60.911 30.000 10.011 40.721 50.764 90.520 70.000 10.442 20.405 80.000 10.667 10.655 30.473 60.614 40.437 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
ODIN - Ins200permissive0.381 60.507 60.375 40.237 50.484 80.108 30.500 10.000 30.125 20.000 10.058 70.647 20.385 60.667 80.853 60.542 41.000 10.000 71.000 10.093 60.000 50.028 90.274 60.682 40.550 10.000 50.269 40.000 10.000 30.714 10.566 21.000 10.000 10.500 10.125 10.585 70.066 70.653 100.083 10.049 20.264 30.227 40.667 20.000 40.000 10.278 100.723 80.250 60.786 91.000 10.744 100.039 60.209 50.494 90.000 40.250 50.446 70.500 50.750 10.200 70.780 50.333 50.602 40.469 50.163 30.406 60.530 80.000 40.668 10.200 40.000 20.000 70.500 10.313 30.769 51.000 10.511 60.196 60.286 60.393 90.337 60.000 20.000 10.600 10.000 70.174 70.226 20.000 40.579 20.200 70.887 10.750 10.428 60.782 70.438 10.000 10.795 50.063 70.003 60.500 30.000 50.333 21.000 10.742 60.083 50.585 30.417 80.448 100.496 60.055 50.734 40.472 20.174 90.000 20.250 50.000 40.688 40.000 31.000 10.000 20.631 60.667 70.275 20.694 101.000 10.000 50.328 20.422 50.000 81.000 10.500 80.638 70.000 50.391 60.000 30.582 70.800 50.208 90.000 60.246 20.667 70.000 50.638 20.167 10.000 30.000 20.778 20.000 20.000 10.563 30.614 60.841 10.333 60.250 60.938 90.569 10.500 80.695 40.264 50.863 40.000 10.000 60.550 91.000 10.668 40.000 10.000 70.667 60.000 10.333 80.333 50.665 10.434 70.264 6
TD3D Scannet200permissive0.320 70.501 70.264 70.164 70.506 70.062 60.500 10.000 30.000 50.000 10.208 30.431 70.252 81.000 10.733 80.587 30.000 70.008 60.000 70.106 50.000 50.356 30.123 90.686 30.101 70.000 50.152 70.000 10.000 30.226 60.280 80.000 70.000 10.250 50.000 30.619 60.061 80.841 30.000 40.000 70.167 60.194 50.333 50.000 40.000 10.667 60.820 30.250 60.790 81.000 10.879 30.077 50.094 80.708 20.217 20.049 70.634 50.792 20.331 90.033 100.716 70.159 70.396 70.331 70.099 60.415 50.842 20.000 40.458 40.542 10.000 20.101 60.000 50.218 50.513 70.500 70.458 70.104 70.516 40.456 50.268 90.000 20.000 10.400 30.022 50.233 50.143 30.000 40.677 10.400 30.504 100.095 80.083 100.890 60.061 30.000 10.906 10.076 60.231 20.125 70.000 50.003 70.792 80.881 10.000 70.098 80.125 90.498 90.459 70.063 30.715 50.000 70.241 80.000 20.396 40.063 20.605 60.000 30.000 70.000 20.448 100.629 80.202 60.967 10.250 50.038 10.192 60.185 70.083 71.000 11.000 10.857 20.000 50.470 50.012 10.565 80.798 60.621 10.111 40.500 11.000 10.017 30.509 50.000 20.008 21.000 10.525 70.000 20.000 10.332 80.679 30.264 70.333 60.267 51.000 10.549 30.299 100.387 70.328 30.744 90.000 10.000 60.435 101.000 10.283 90.000 10.196 30.817 30.000 10.472 40.222 70.123 90.560 60.156 7
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
Mask3D Scannet2000.388 50.542 50.357 60.237 60.610 50.091 40.125 100.000 30.000 50.000 10.065 60.668 10.451 41.000 10.955 40.640 20.500 40.039 30.125 60.063 70.409 40.311 40.291 50.609 60.266 40.000 50.163 60.000 10.008 20.044 70.496 41.000 10.000 10.018 70.000 30.756 50.573 20.808 40.000 40.010 60.042 80.130 80.552 30.042 20.000 11.000 10.725 70.750 10.883 11.000 10.832 60.024 70.107 60.614 70.226 10.250 50.628 60.792 20.677 50.400 10.741 60.278 60.511 60.077 100.111 40.313 70.715 60.302 10.017 80.200 40.000 20.188 50.000 50.178 70.736 61.000 10.615 20.514 40.409 50.380 100.600 20.000 20.000 10.400 30.013 60.254 40.381 10.000 40.123 80.400 30.839 30.258 60.463 50.926 40.265 20.000 10.857 20.099 40.021 40.500 30.027 30.028 51.000 10.502 100.016 60.076 90.500 20.612 50.578 50.005 70.597 60.194 50.497 50.000 20.500 10.000 40.323 90.000 31.000 10.000 20.748 30.708 50.050 90.890 51.000 10.008 20.151 80.301 61.000 11.000 10.792 60.945 11.000 10.511 40.004 20.753 40.776 70.287 50.020 50.003 90.974 30.033 10.412 100.000 20.000 30.000 20.667 50.000 20.000 10.491 60.676 40.352 60.335 50.060 70.822 100.527 41.000 10.517 60.606 10.853 50.000 10.004 50.806 11.000 10.727 30.000 10.042 60.739 50.000 10.399 70.391 40.504 50.591 50.571 2
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
Minkowski 34D Inst.permissive0.203 100.369 90.134 100.078 100.479 90.003 90.500 10.000 30.000 50.000 10.100 50.371 80.300 70.667 80.746 70.400 60.000 70.000 70.000 70.031 80.000 50.074 80.165 80.413 100.000 90.000 50.070 90.000 10.000 30.000 80.221 100.000 70.000 10.000 80.000 30.372 100.070 60.706 80.000 40.000 70.000 100.123 90.033 100.000 40.000 10.422 90.732 60.000 90.778 101.000 10.845 50.000 80.090 90.636 60.000 40.000 80.158 90.000 70.250 100.050 90.693 80.123 90.051 100.385 60.009 90.118 100.406 100.000 40.000 90.200 40.000 20.000 70.000 50.133 90.307 100.500 70.251 90.000 90.281 70.402 80.317 70.000 20.000 10.000 80.000 70.060 90.000 50.000 40.396 50.200 70.669 70.021 90.218 90.720 100.000 40.000 10.696 80.025 90.000 80.000 80.000 50.000 80.125 100.596 70.000 70.191 60.500 20.595 60.369 90.000 80.500 90.000 70.143 100.000 20.000 80.000 40.226 100.000 30.000 70.000 20.701 50.511 90.000 100.851 80.000 70.000 50.150 90.052 100.100 60.981 70.500 80.286 80.000 50.000 100.000 30.545 90.522 100.250 70.000 60.000 100.522 100.000 50.500 60.000 20.000 30.000 20.282 100.000 20.000 10.178 100.382 90.018 100.056 90.000 80.997 40.107 100.677 60.313 90.000 90.726 100.000 10.000 60.583 80.903 80.200 100.000 10.000 70.333 90.000 10.442 60.083 80.109 100.387 90.000 10
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019
CSC-Pretrain Inst.permissive0.209 90.361 100.157 90.085 90.506 60.007 80.500 10.000 30.000 50.000 10.000 100.093 100.221 90.667 80.524 100.400 60.000 70.000 70.000 70.004 90.000 50.000 100.109 100.589 80.000 90.000 50.059 100.000 10.000 30.000 80.322 60.000 70.000 10.000 80.000 30.405 80.055 90.700 90.000 40.000 70.028 90.091 100.083 70.000 40.000 10.667 60.768 50.000 90.807 61.000 10.776 90.000 80.000 100.340 100.000 40.000 80.103 100.000 70.750 10.200 70.634 100.053 100.246 80.677 30.006 100.198 80.432 90.000 40.000 90.050 90.000 20.000 70.000 50.111 100.356 90.500 70.188 100.000 90.220 80.448 60.050 100.000 20.000 10.000 80.000 70.032 100.000 50.000 40.396 50.000 90.573 90.000 100.228 80.747 90.000 40.000 10.573 100.021 100.000 80.000 80.000 50.000 80.500 90.573 80.000 70.000 100.125 90.592 70.364 100.000 80.450 100.000 70.364 60.000 20.000 80.000 40.340 80.000 30.000 70.000 20.610 80.833 40.221 40.702 90.000 70.000 50.135 100.094 90.125 40.571 90.500 80.143 100.000 50.125 80.000 30.618 60.667 90.115 100.000 60.125 61.000 10.000 50.500 60.000 20.000 30.000 20.502 90.000 20.000 10.312 90.248 100.050 90.000 100.000 80.997 40.420 70.500 80.149 100.451 20.748 70.000 10.000 60.636 70.667 100.600 60.000 10.000 70.278 100.000 10.333 80.000 100.294 70.381 100.110 8
Ji Hou, Benjamin Graham, Matthias Nießner, Saining Xie: Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts. CVPR 2021
LGround Inst.permissive0.246 80.413 80.170 80.130 80.455 100.003 100.500 10.000 30.000 50.000 10.017 90.333 90.111 101.000 10.681 90.400 60.000 70.000 71.000 10.003 100.000 50.167 70.190 70.637 50.067 80.000 50.081 80.000 10.000 30.000 80.264 90.000 70.000 10.000 80.000 30.387 90.031 100.754 70.000 40.000 70.151 70.135 70.056 90.000 40.000 10.582 80.589 100.500 20.815 51.000 10.903 10.000 80.097 70.588 80.000 40.000 80.234 80.000 70.500 70.400 10.682 90.156 80.159 90.750 10.046 80.125 90.660 70.000 40.200 60.000 100.000 20.000 70.000 50.164 80.402 80.500 70.373 80.025 80.143 100.426 70.317 70.000 20.000 10.000 80.000 70.063 80.000 50.000 40.000 90.000 90.575 80.250 70.241 70.772 80.000 40.000 10.653 90.034 80.000 80.000 80.000 50.000 81.000 10.561 90.000 70.100 70.500 20.541 80.452 80.000 80.581 70.000 70.364 60.000 20.000 80.000 40.571 70.000 30.000 70.000 20.568 90.511 90.167 70.857 70.000 70.000 50.164 70.112 80.000 80.530 101.000 10.286 80.000 50.125 80.000 30.464 100.706 80.208 80.000 60.125 60.744 40.000 50.500 60.000 20.000 30.000 20.511 80.000 20.000 10.344 70.541 80.068 80.333 60.000 81.000 10.196 90.533 70.318 80.000 90.748 80.000 10.000 60.690 61.000 10.400 80.000 10.000 70.667 60.000 10.333 80.333 50.270 80.399 80.083 9
David Rozenberszki, Or Litany, Angela Dai: Language-Grounded Indoor 3D Semantic Segmentation in the Wild.