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 bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort 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 by
Minkowski 34D Inst.permissive0.280 40.488 40.192 50.124 40.593 40.010 40.500 10.000 10.000 10.000 10.447 40.535 40.445 31.000 10.861 40.400 30.225 20.000 30.000 30.142 40.000 20.074 40.342 30.467 50.067 30.000 10.119 50.000 10.000 20.000 40.337 50.000 20.000 10.000 40.000 10.506 50.070 20.804 40.000 30.000 40.333 30.172 30.150 50.000 20.000 10.479 50.745 30.000 50.830 51.000 10.904 30.167 20.090 40.732 20.000 30.000 30.443 40.000 30.500 30.542 10.772 50.396 40.077 50.385 40.044 40.118 50.777 40.000 20.000 40.200 30.000 10.000 30.000 10.148 40.502 40.500 20.419 40.159 50.281 40.404 50.317 30.000 10.000 10.200 30.000 40.077 30.000 30.000 10.750 30.200 30.715 40.021 40.551 20.828 50.000 30.000 10.743 40.059 50.000 30.000 30.000 20.000 30.125 50.648 30.000 20.191 20.500 10.669 40.502 40.000 50.568 40.000 20.516 40.000 10.000 30.000 20.305 50.000 10.000 20.000 10.825 10.833 20.021 50.918 20.000 30.000 30.191 40.346 40.100 40.981 31.000 10.286 40.000 20.000 50.000 30.868 40.648 50.292 30.000 30.375 31.000 10.000 30.500 20.000 40.333 10.000 20.538 50.000 10.000 10.213 50.518 40.098 40.528 10.250 30.997 30.284 50.677 30.398 30.167 40.790 40.000 10.000 20.618 50.903 50.200 50.000 10.333 10.333 40.000 10.442 30.083 40.213 40.587 40.131 5
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019
TD3D Scannet200permissive0.379 20.603 20.306 20.190 20.635 20.073 20.500 10.000 10.000 10.000 10.495 30.735 20.275 51.000 10.979 20.590 20.000 40.021 20.000 30.146 30.000 20.356 20.173 50.795 10.226 20.000 10.173 20.000 10.000 20.226 20.390 20.000 20.000 10.250 10.000 10.706 20.061 30.885 10.093 20.186 20.259 40.200 10.667 10.000 20.000 10.667 20.825 10.250 40.834 41.000 10.958 10.553 10.111 30.748 10.220 20.051 20.866 20.792 10.390 50.045 50.800 20.302 50.517 10.533 30.113 20.427 10.843 20.000 20.458 10.600 10.000 10.101 20.000 10.259 10.717 20.500 20.615 20.520 20.526 20.457 10.270 40.000 10.000 10.400 20.088 20.294 20.181 20.000 11.000 10.400 10.710 50.103 30.477 50.905 20.061 20.000 10.906 20.102 20.232 10.125 20.000 20.003 20.792 31.000 10.000 20.102 30.125 40.559 50.523 30.075 20.715 10.000 20.424 50.000 10.396 20.250 10.638 10.000 10.000 20.000 10.622 50.833 20.221 10.970 10.250 20.038 10.260 20.415 10.125 21.000 11.000 10.857 20.000 20.908 10.012 10.869 30.836 10.635 10.111 10.625 11.000 10.020 20.510 10.003 30.009 21.000 10.778 10.000 10.000 10.370 30.755 10.288 20.333 30.274 21.000 10.557 10.731 20.456 20.433 30.769 50.000 10.000 20.621 41.000 10.458 40.000 10.196 20.817 10.000 10.472 10.222 30.205 50.689 20.274 3
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
Mask3D Scannet2000.445 10.653 10.392 10.254 10.648 10.097 10.125 50.000 10.000 10.000 10.657 10.971 10.451 21.000 11.000 10.640 10.500 10.045 11.000 10.241 20.409 10.363 10.440 10.686 30.300 10.000 10.201 10.000 10.009 10.290 10.556 11.000 10.000 10.063 30.000 10.830 10.573 10.844 20.333 10.204 10.058 50.158 50.552 20.056 10.000 11.000 10.725 40.750 10.927 11.000 10.888 40.042 30.120 20.615 40.226 10.250 10.890 10.792 10.677 20.510 20.818 10.699 10.512 20.167 50.125 10.315 20.943 10.309 10.017 30.200 30.000 10.188 10.000 10.183 30.815 11.000 10.827 10.741 10.442 30.414 40.600 10.000 10.000 10.458 10.049 30.321 10.381 10.000 10.908 20.400 10.841 10.260 10.710 10.966 10.265 10.000 10.924 10.152 10.025 20.500 10.027 10.028 11.000 10.556 50.016 10.080 50.500 10.694 30.608 10.084 10.604 30.194 10.538 30.000 10.500 10.000 20.354 40.000 11.000 10.000 10.761 20.930 10.053 40.890 31.000 10.008 20.262 10.358 21.000 11.000 10.792 40.966 11.000 10.765 20.004 20.930 10.780 20.330 20.027 20.625 10.974 40.050 10.412 50.021 20.000 30.000 20.778 10.000 10.000 10.493 20.746 20.454 10.335 20.396 10.930 50.551 21.000 10.552 10.606 10.853 10.000 10.004 10.806 11.000 10.727 20.000 10.042 30.745 20.000 10.399 40.391 10.630 10.721 10.619 1
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
CSC-Pretrain Inst.permissive0.275 50.466 50.218 40.110 50.625 30.007 50.500 10.000 10.000 10.000 10.000 50.222 50.377 41.000 10.661 50.400 30.000 40.000 30.000 30.119 50.000 20.000 50.277 40.685 40.067 30.000 10.132 30.000 10.000 20.000 40.367 40.000 20.000 10.000 40.000 10.591 30.055 40.783 50.000 30.014 30.500 20.161 40.278 30.000 20.000 10.667 20.768 20.500 20.866 21.000 10.829 50.000 40.019 50.555 50.000 30.000 30.305 50.000 30.750 10.200 40.783 40.429 30.395 30.677 20.020 50.286 30.584 50.000 20.000 40.115 50.000 10.000 30.000 10.145 50.423 50.500 20.364 50.369 40.571 10.448 30.206 50.000 10.000 10.200 30.106 10.065 50.000 30.000 10.750 30.200 30.774 20.000 50.501 30.841 40.000 30.000 10.692 50.063 40.000 30.000 30.000 20.000 30.500 40.649 20.000 20.084 40.125 40.719 10.413 50.004 40.450 50.000 20.638 10.000 10.000 30.000 20.505 30.000 10.000 20.000 10.727 30.833 20.221 20.779 50.000 30.000 30.168 50.311 50.125 20.571 40.500 50.143 50.000 20.250 40.000 30.869 20.667 40.162 50.000 30.250 41.000 10.000 30.500 20.000 40.000 30.000 20.689 40.000 10.000 10.312 40.383 50.114 30.333 30.000 40.997 30.420 30.613 40.212 50.500 20.819 20.000 10.000 20.768 21.000 10.918 10.000 10.000 40.278 50.000 10.333 50.000 50.353 20.546 50.258 4
Ji Hou, Benjamin Graham, Matthias Nießner, Saining Xie: Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts. CVPR 2021
LGround Inst.permissive0.314 30.529 30.225 30.155 30.578 50.010 30.500 10.000 10.000 10.000 10.515 20.556 30.696 11.000 10.927 30.400 30.083 30.000 31.000 10.252 10.000 20.167 30.350 20.731 20.067 30.000 10.123 40.000 10.000 20.036 30.372 30.000 20.000 10.250 10.000 10.569 40.031 50.810 30.000 30.000 40.630 10.183 20.278 30.000 20.000 10.582 40.589 50.500 20.863 31.000 10.940 20.000 40.144 10.716 30.000 30.000 30.484 30.000 30.500 30.400 30.798 30.500 20.278 40.750 10.093 30.166 40.783 30.000 20.200 20.400 20.000 10.000 30.000 10.219 20.539 30.500 20.578 30.413 30.181 50.457 20.375 20.000 10.000 10.050 50.000 40.077 40.000 30.000 10.500 50.000 50.743 30.250 20.488 40.846 30.000 30.000 10.800 30.069 30.000 30.000 30.000 20.000 31.000 10.607 40.000 20.200 10.500 10.694 20.528 20.063 30.659 20.000 20.594 20.000 10.000 30.000 20.571 20.000 10.000 20.000 10.716 40.647 50.221 20.857 40.000 30.000 30.217 30.346 30.071 50.530 51.000 10.429 30.000 20.286 30.000 30.826 50.706 30.208 40.000 30.250 40.744 50.000 30.500 20.042 10.000 30.000 20.746 30.000 10.000 10.517 10.625 30.085 50.333 30.000 41.000 10.378 40.533 50.376 40.042 50.814 30.000 10.000 20.765 31.000 10.600 30.000 10.000 40.667 30.000 10.472 10.333 20.337 30.605 30.305 2
David Rozenberszki, Or Litany, Angela Dai: Language-Grounded Indoor 3D Semantic Segmentation in the Wild.