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%chairtabledoorcouchcabinetshelfdeskoffice chairbedpillowsinkpicturewindowtoiletbookshelfmonitorcurtainbookarmchaircoffee tableboxrefrigeratorlampkitchen cabinettowelclothestvnightstandcounterdresserstoolcushionplantceilingbathtubend tabledining tablekeyboardbagbackpacktoilet paperprintertv standwhiteboardblanketshower curtaintrash canclosetstairsmicrowavestoveshoecomputer towerbottlebinottomanbenchboardwashing machinemirrorcopierbasketsofa chairfile cabinetfanlaptopshowerpaperpersonpaper towel dispenserovenblindsrackplateblackboardpianosuitcaserailradiatorrecycling bincontainerwardrobesoap dispensertelephonebucketclockstandlightlaundry basketpipeclothes dryerguitartoilet paper holderseatspeakercolumnbicycleladderbathroom stallshower wallcupjacketstorage bincoffee makerdishwasherpaper towel rollmachinematwindowsillbartoasterbulletin boardironing boardfireplacesoap dishkitchen counterdoorframetoilet paper dispensermini fridgefire extinguisherballhatshower curtain rodwater coolerpaper cuttertrayshower doorpillarledgetoaster ovenmousetoilet seat cover dispenserfurniturecartstorage containerscaletissue boxlight switchcratepower outletdecorationsignprojectorcloset doorvacuum cleanercandleplungerstuffed animalheadphonesdish rackbroomguitar caserange hooddustpanhair dryerwater bottlehandicap barpurseventshower floorwater pitchermailboxbowlpaper bagalarm clockmusic standprojector screendividerlaundry detergentbathroom counterobjectbathroom vanitycloset walllaundry hamperbathroom stall doorceiling lighttrash bindumbbellstair railtubebathroom cabinetcd casecloset rodcoffee kettlestructureshower headkeyboard pianocase of water bottlescoat rackstorage organizerfolded chairfire alarmpower stripcalendarposterpotted plantluggagemattress
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 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 by
DINO3D-Scannet200copyleft0.511 40.685 40.484 10.331 30.892 30.821 20.890 30.907 60.629 20.468 20.905 10.001 81.000 10.816 10.968 20.863 40.811 50.944 90.596 100.960 60.778 40.532 30.719 100.481 10.851 90.803 10.873 10.850 10.421 10.806 50.856 70.111 60.761 30.677 10.000 50.944 10.861 51.000 10.220 20.708 40.856 60.220 10.864 10.579 11.000 10.764 110.655 40.327 51.000 10.911 40.244 10.667 100.923 10.857 10.702 10.889 40.496 20.048 20.355 110.494 20.794 50.798 31.000 10.042 50.264 80.817 70.683 20.675 10.167 50.000 60.700 10.824 40.417 70.000 60.000 50.764 10.000 80.500 30.699 30.789 70.079 90.472 20.845 40.930 10.000 40.667 10.000 50.412 30.000 20.163 61.000 10.000 60.419 10.500 21.000 10.777 20.576 30.867 40.378 20.334 50.028 30.764 40.542 20.559 10.000 50.800 10.528 50.000 40.346 60.714 10.125 50.756 50.754 60.866 50.750 10.600 30.500 10.500 11.000 10.667 11.000 10.000 10.298 10.000 50.250 60.194 20.000 80.850 40.000 50.250 50.595 10.000 30.063 10.860 50.000 10.714 20.000 10.944 10.750 10.000 10.974 20.000 10.000 10.857 40.655 40.719 100.250 30.014 70.000 11.000 10.000 10.142 40.744 30.200 80.746 30.436 30.221 70.798 10.500 80.011 50.000 10.385 80.000 10.000 20.000 60.792 70.663 10.000 60.000 20.200 50.000 50.000 41.000 10.000 2
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
AQ3D-ScanNet2000.527 30.743 10.477 30.324 40.930 10.784 40.886 40.947 30.631 10.410 40.893 20.328 20.955 70.795 20.926 50.925 20.816 31.000 10.754 40.960 50.858 10.556 10.805 50.396 30.929 30.707 30.839 40.835 30.360 20.764 90.896 50.087 80.778 20.576 30.160 30.879 20.901 31.000 10.089 80.023 70.895 40.161 30.862 20.549 60.667 61.000 10.546 70.435 11.000 10.935 20.113 71.000 10.831 60.778 20.602 40.938 10.279 60.025 50.567 40.458 31.000 10.750 41.000 10.833 10.392 30.991 10.479 50.434 61.000 10.071 20.428 50.868 21.000 10.333 50.281 10.000 70.533 61.000 10.671 40.838 50.534 20.664 10.819 60.768 30.083 10.009 80.000 50.465 10.000 20.337 20.667 40.030 20.172 30.250 71.000 10.738 40.513 60.712 80.368 30.366 40.016 50.489 90.600 10.083 60.563 20.400 30.923 10.500 10.736 20.607 30.500 10.837 20.843 21.000 10.658 40.318 60.500 10.000 21.000 10.333 91.000 10.000 10.105 50.000 50.500 10.000 41.000 11.000 10.000 50.141 80.026 80.000 30.000 31.000 10.000 10.714 20.000 10.621 50.750 10.000 11.000 10.000 10.000 10.714 80.667 31.000 10.167 50.667 10.000 11.000 10.000 10.088 80.873 10.517 30.556 70.073 70.434 30.458 70.707 30.083 20.000 10.803 30.000 10.000 21.000 11.000 10.000 40.056 30.000 20.200 50.143 30.000 40.250 70.000 2
ODIN - Ins200permissive0.451 60.637 70.407 50.277 60.742 110.699 80.855 60.826 110.626 30.441 30.742 80.003 70.941 80.637 60.910 70.616 100.679 80.944 90.695 70.877 80.763 50.357 70.723 90.475 20.779 100.494 60.782 70.795 50.334 30.824 30.867 60.108 70.701 50.638 20.000 50.873 30.749 70.667 110.203 30.500 50.886 50.116 40.583 100.571 20.688 51.000 10.760 10.162 81.000 10.852 60.078 80.833 60.887 20.778 20.577 60.859 90.550 10.000 80.542 50.028 100.667 80.874 11.000 10.125 30.232 90.870 40.406 70.337 80.167 50.000 60.671 30.742 70.500 40.000 60.000 50.528 21.000 10.417 90.597 50.872 20.275 30.000 90.800 70.850 20.000 40.528 20.000 50.215 80.000 20.238 50.667 40.000 60.019 70.250 71.000 10.429 80.599 20.778 50.221 60.370 30.284 10.278 110.400 80.125 40.000 50.200 80.404 70.000 40.250 80.714 10.500 10.504 80.769 50.677 80.750 10.963 10.500 10.000 20.500 100.333 91.000 10.000 10.000 90.438 10.500 10.000 41.000 10.333 80.226 20.250 50.250 50.000 30.000 30.668 80.000 10.494 100.000 10.000 80.750 10.000 10.833 70.000 10.000 10.777 70.333 60.944 40.000 60.333 30.000 11.000 10.000 10.089 70.407 90.600 10.823 20.080 60.264 60.469 60.717 20.000 70.000 10.500 60.000 10.000 20.000 61.000 10.125 20.333 10.000 20.200 50.000 50.000 41.000 10.000 2
Volt-SPFormerpermissive0.527 20.731 30.475 40.342 20.826 60.803 30.942 10.950 20.594 40.321 50.867 50.008 60.994 50.767 30.926 60.874 30.815 41.000 10.810 20.973 30.856 30.510 50.825 20.346 50.923 40.799 20.843 30.812 40.262 50.923 10.921 30.279 20.901 10.500 70.000 50.801 40.937 11.000 10.329 10.000 80.903 20.076 60.789 30.565 30.907 21.000 10.614 60.413 31.000 10.937 10.214 20.629 110.878 30.725 80.579 50.880 50.433 30.020 70.400 70.547 11.000 10.843 21.000 10.125 30.343 40.855 50.750 10.449 51.000 10.057 30.700 10.802 50.500 40.850 20.011 30.047 61.000 11.000 10.715 20.875 10.255 40.099 70.857 20.738 50.000 40.056 70.025 20.372 40.250 10.279 40.667 40.002 50.000 80.250 70.500 51.000 10.391 90.737 70.309 50.397 20.000 80.817 30.542 20.557 21.000 10.400 30.681 30.000 40.500 40.519 60.500 10.773 40.818 40.884 40.656 50.510 40.500 10.000 21.000 10.472 41.000 10.000 10.027 80.000 50.331 50.000 41.000 11.000 10.000 50.500 10.304 30.000 30.000 31.000 10.000 10.714 20.000 10.677 20.750 10.000 10.944 40.000 10.000 11.000 10.764 20.833 70.250 30.278 40.000 11.000 10.000 10.103 50.753 20.600 10.508 100.638 10.167 90.458 70.741 10.019 40.000 10.850 20.000 10.000 21.000 11.000 10.000 40.028 50.000 20.200 50.000 50.250 11.000 10.000 2
Kadir Yilmaz, Adrian Kruse, Tristan Höfer, Daan de Geus, Bastian Leibe: Volume Transformer: Revisiting Vanilla Transformers for 3D Scene Understanding.
ACGP-ScanNet2000.544 10.737 20.483 20.381 10.801 90.859 10.921 20.912 50.536 60.483 10.846 70.036 40.996 40.699 50.955 30.929 10.842 21.000 10.834 10.993 10.858 20.517 40.838 10.396 40.968 20.682 40.860 20.840 20.292 40.800 60.825 80.213 40.573 60.552 40.000 50.738 50.918 21.000 10.064 91.000 10.897 30.184 20.747 40.424 70.835 31.000 10.718 20.388 41.000 10.915 30.140 60.721 80.851 40.778 20.666 20.902 30.405 40.035 40.511 60.307 60.903 30.680 71.000 10.708 20.403 20.931 30.658 30.510 31.000 10.089 10.671 30.765 61.000 10.528 30.006 40.204 41.000 10.500 30.759 10.854 30.590 10.461 30.850 30.767 40.042 20.086 50.000 50.462 20.000 20.349 11.000 10.007 40.341 20.444 61.000 10.759 30.371 100.867 30.367 40.462 10.000 80.903 20.443 70.500 30.250 30.600 20.809 20.500 10.944 10.540 50.000 60.944 10.905 10.944 30.677 30.637 20.500 10.000 21.000 10.507 31.000 10.000 10.140 30.000 50.500 10.000 41.000 11.000 10.143 40.146 70.396 20.000 30.000 31.000 10.000 10.782 10.000 10.638 40.677 60.000 10.974 20.000 10.000 10.959 30.903 10.884 61.000 10.472 20.000 10.250 70.000 10.185 10.718 40.391 60.604 60.189 50.206 80.500 50.637 40.064 30.000 10.667 40.000 10.000 21.000 11.000 10.050 30.000 60.000 20.317 40.144 20.024 31.000 10.008 1
Rongkun Yang, Ye Zhang, Longguang Wang, Zhiheng Fu, Lian Xu, Yulan Guo: Beyond Context Bias: Adaptive Instance Placement for Robust 3D Instance Segmentation.
TD3D Scannet200permissive0.379 80.603 80.306 80.190 80.885 40.755 50.800 80.958 10.390 70.260 80.866 60.232 30.979 60.523 90.869 90.559 110.689 71.000 10.795 30.905 70.748 70.173 110.825 30.173 80.970 10.457 70.615 80.456 80.200 70.621 100.906 40.553 10.517 70.510 60.220 20.715 60.706 81.000 10.113 60.792 20.717 80.073 70.635 70.557 40.638 71.000 10.205 110.146 91.000 10.769 110.186 51.000 10.710 110.778 20.415 70.834 100.226 80.021 60.590 30.356 50.817 40.477 111.000 10.000 60.635 10.843 60.427 60.270 100.125 70.000 60.102 91.000 10.125 80.000 60.000 50.000 70.000 80.125 100.370 90.622 110.221 50.196 50.836 50.288 80.000 40.093 40.020 30.294 60.000 20.075 80.667 40.038 10.111 50.250 70.000 100.526 60.495 70.908 10.111 90.259 60.003 70.667 60.045 110.000 80.000 50.400 30.274 90.000 40.274 70.226 80.000 60.520 70.302 110.731 70.103 90.458 50.500 10.000 21.000 10.472 40.792 90.000 10.088 60.061 40.250 60.009 30.250 60.333 80.181 30.396 40.051 70.012 10.000 30.458 100.000 10.424 110.000 10.101 70.390 110.000 10.833 70.000 10.000 10.857 40.222 81.000 10.000 60.003 80.000 10.000 80.000 10.102 60.275 110.400 50.735 40.061 90.433 40.533 40.625 50.000 70.000 10.259 100.000 10.000 20.000 60.500 80.000 40.000 61.000 10.600 10.000 50.250 10.000 80.000 2
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
LGround Inst.permissive0.314 90.529 90.225 90.155 90.810 70.625 90.798 90.940 40.372 90.217 90.484 90.000 90.927 90.528 80.826 110.694 70.605 91.000 10.731 50.846 90.716 90.350 80.589 110.123 100.857 80.457 80.578 90.376 100.183 80.765 80.800 90.000 100.278 100.500 70.000 50.659 70.569 101.000 10.093 70.000 80.539 90.010 90.578 110.378 100.571 81.000 10.337 90.252 60.530 110.814 90.000 100.744 70.743 90.746 70.346 90.863 80.067 90.000 80.400 70.167 80.667 80.488 101.000 10.000 60.208 100.783 80.166 100.375 70.071 100.000 60.200 70.607 100.000 90.000 60.000 50.000 71.000 10.500 30.517 60.716 100.221 60.000 90.706 90.085 110.000 40.000 90.000 50.077 100.000 20.063 90.278 90.000 60.000 80.500 20.083 90.181 110.515 50.286 90.144 70.219 80.042 20.582 80.400 80.000 80.000 50.000 110.305 80.000 40.000 100.036 90.000 60.413 90.500 80.533 110.250 80.200 70.500 10.000 21.000 10.472 41.000 10.000 10.000 90.000 50.250 60.000 40.000 80.333 80.000 50.000 90.000 90.000 30.000 30.600 90.000 10.594 60.000 10.000 80.500 80.000 10.647 110.000 10.000 10.429 90.333 60.500 110.000 60.000 90.000 10.000 80.000 10.069 90.696 50.050 110.556 70.031 110.042 110.750 20.250 100.000 70.000 10.630 50.000 10.000 20.000 60.500 80.000 40.000 60.000 20.400 20.000 50.000 40.000 80.000 2
David Rozenberszki, Or Litany, Angela Dai: Language-Grounded Indoor 3D Semantic Segmentation in the Wild.
Mask3D Scannet2000.445 70.653 60.392 70.254 70.844 50.746 60.818 70.888 80.556 50.262 70.890 30.025 51.000 10.608 70.930 40.694 80.721 60.930 110.686 80.966 40.615 100.440 60.725 80.201 70.890 70.414 100.827 50.552 70.158 110.806 40.924 20.042 90.512 80.412 110.226 10.604 80.830 61.000 10.125 40.792 20.815 70.097 50.648 60.551 50.354 101.000 10.630 50.241 71.000 10.853 50.204 40.974 50.841 50.778 20.358 80.927 20.300 50.045 30.640 20.363 40.745 70.710 51.000 10.000 60.330 50.943 20.315 80.600 21.000 10.027 40.080 110.556 110.500 40.409 40.000 50.194 51.000 10.500 30.493 70.761 80.053 100.042 80.780 80.454 70.009 30.333 30.050 10.321 50.000 20.084 70.552 80.008 30.027 60.750 10.500 50.442 70.657 10.765 60.120 80.183 90.021 41.000 10.510 60.016 70.000 50.400 30.619 40.000 40.396 50.290 70.000 60.741 60.699 71.000 10.260 70.017 90.125 110.000 20.792 90.399 81.000 10.000 10.049 70.265 30.063 90.000 41.000 10.335 70.381 10.500 10.250 50.004 20.000 30.727 70.000 10.538 80.000 10.188 60.677 60.000 10.930 50.000 10.000 10.966 20.391 50.908 50.000 60.028 60.000 11.000 10.000 10.152 30.451 60.458 40.971 10.573 20.606 10.167 110.625 50.004 60.000 10.058 110.000 10.000 21.000 11.000 10.000 40.056 30.000 20.200 50.309 10.000 41.000 10.000 2
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
Minkowski 34D Inst.permissive0.280 100.488 100.192 110.124 100.804 80.518 100.772 110.904 70.337 110.191 100.443 100.000 90.861 100.502 100.868 100.669 90.587 100.997 70.467 110.828 110.732 80.342 90.745 70.119 110.918 50.404 110.419 100.398 90.172 90.618 110.743 100.167 50.077 110.500 70.000 50.568 90.506 111.000 10.044 100.000 80.502 100.010 100.593 90.284 110.305 110.903 100.213 100.142 100.981 80.790 100.000 101.000 10.715 100.538 110.346 100.830 110.067 90.000 80.400 70.074 90.333 100.551 81.000 10.000 60.292 70.777 90.118 110.317 90.100 90.000 60.191 80.648 90.000 90.000 60.000 50.000 70.000 80.500 30.213 110.825 60.021 110.333 40.648 110.098 100.000 40.000 90.000 50.077 90.000 20.000 110.150 110.000 60.000 80.000 110.225 80.281 100.447 80.000 110.090 100.148 100.000 80.479 100.542 20.000 80.000 50.200 80.131 110.000 40.250 80.000 100.000 60.159 110.396 100.677 80.021 100.000 100.500 10.000 21.000 10.442 70.125 110.000 10.000 90.000 50.000 100.333 10.000 80.528 50.000 50.000 90.000 90.000 30.000 30.200 110.000 10.516 90.000 10.000 80.500 80.000 10.833 70.000 10.000 10.286 100.083 100.750 80.000 60.000 90.000 10.000 80.000 10.059 110.445 70.200 80.535 90.070 80.167 90.385 90.375 90.000 70.000 10.333 90.000 10.000 20.000 60.500 80.000 40.000 60.000 20.200 50.000 50.000 40.000 80.000 2
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019
CompetitorFormer-2000.469 50.676 50.401 60.296 50.901 20.729 70.885 50.829 90.380 80.320 60.873 40.400 10.998 30.711 40.980 10.847 50.854 11.000 10.696 60.989 20.759 60.556 20.806 40.240 60.918 50.650 50.818 60.629 60.224 60.839 20.933 10.247 30.711 40.540 50.021 40.543 100.900 40.903 100.118 50.125 60.916 10.057 80.692 50.410 90.747 41.000 10.664 30.424 20.933 90.839 70.207 30.703 90.748 80.700 90.610 30.869 60.270 70.068 10.878 10.244 70.794 50.698 61.000 10.000 60.325 60.770 100.482 40.452 40.025 110.015 50.293 60.829 30.663 31.000 10.013 20.385 30.250 70.500 30.491 80.850 40.214 80.131 60.878 10.617 60.000 40.085 60.009 40.278 70.000 20.295 31.000 10.000 60.160 40.500 20.500 50.342 90.534 40.901 20.474 10.222 70.011 60.724 50.542 20.125 40.083 40.336 70.500 60.083 30.565 30.587 40.500 10.827 30.829 30.750 60.508 60.018 80.500 10.000 21.000 10.667 11.000 10.000 10.173 20.286 20.500 10.000 40.125 70.489 60.000 50.500 10.269 40.000 30.050 20.834 60.000 10.581 70.000 10.677 20.467 100.000 10.886 60.000 10.000 10.820 60.144 91.000 11.000 10.103 50.000 11.000 10.000 10.175 20.410 80.330 70.701 50.257 40.292 50.285 100.574 70.157 10.000 10.863 10.000 10.056 10.250 51.000 10.000 40.109 20.000 20.400 20.025 40.000 41.000 10.000 2
CSC-Pretrain Inst.permissive0.275 110.466 110.218 100.110 110.783 100.383 110.783 100.829 100.367 100.168 110.305 110.000 90.661 110.413 110.869 80.719 60.546 110.997 70.685 90.841 100.555 110.277 100.768 60.132 90.779 100.448 90.364 110.212 110.161 100.768 70.692 110.000 100.395 90.500 70.000 50.450 110.591 91.000 10.020 110.000 80.423 110.007 110.625 80.420 80.505 91.000 10.353 80.119 110.571 100.819 80.014 91.000 10.774 70.689 100.311 110.866 70.067 90.000 80.400 70.000 110.278 110.501 91.000 10.000 60.162 110.584 110.286 90.206 110.125 70.000 60.084 100.649 80.000 90.000 60.000 50.000 70.000 80.125 100.312 100.727 90.221 60.000 90.667 100.114 90.000 40.000 90.000 50.065 110.000 20.004 100.278 90.000 60.000 80.500 20.000 100.571 50.000 110.250 100.019 110.145 110.000 80.667 60.200 100.000 80.000 50.200 80.258 100.000 40.000 100.000 100.000 60.369 100.429 90.613 100.000 110.000 100.500 10.000 20.500 100.333 90.500 100.000 10.106 40.000 50.000 100.000 40.000 80.333 80.000 50.000 90.000 90.000 30.000 30.918 40.000 10.638 50.000 10.000 80.750 10.000 10.833 70.000 10.000 10.143 110.000 110.750 80.000 60.000 90.000 10.000 80.000 10.063 100.377 100.200 80.222 110.055 100.500 20.677 30.250 100.000 70.000 10.500 60.000 10.000 20.000 60.500 80.000 40.000 60.000 20.115 110.000 50.000 40.000 80.000 2
Ji Hou, Benjamin Graham, Matthias Nießner, Saining Xie: Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts. CVPR 2021