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 aphead apcommon aptail apalarm 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 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 by
TD3D Scannet200permissive0.211 70.332 70.177 70.103 70.337 70.036 70.222 90.000 30.000 50.000 10.031 40.342 60.093 90.852 40.452 90.559 30.000 70.004 60.000 70.039 50.000 50.309 30.047 90.380 50.028 70.000 50.080 70.000 10.000 30.147 60.192 80.000 70.000 10.083 60.000 30.395 60.039 90.662 40.000 40.000 70.074 50.135 50.296 60.000 40.000 10.231 90.646 30.139 70.633 71.000 10.705 20.048 50.088 70.439 50.184 20.039 70.266 60.551 40.260 80.026 100.463 70.046 80.252 60.249 40.083 50.372 50.411 50.000 40.414 40.323 10.000 20.052 60.000 50.157 60.278 70.278 70.237 70.015 70.321 50.253 50.060 90.000 20.000 10.272 60.008 50.169 50.032 40.000 40.404 10.356 50.283 70.073 80.028 100.617 60.038 30.000 10.494 50.037 60.215 20.083 70.000 50.003 70.486 70.694 20.000 70.040 90.083 90.219 100.209 70.007 60.483 50.000 70.125 80.000 20.150 60.014 20.544 60.000 30.000 70.000 20.260 90.143 100.200 40.610 70.028 60.032 10.145 60.059 70.046 70.740 60.806 30.543 60.000 50.108 70.008 10.222 100.669 60.456 10.074 40.224 10.586 10.006 30.451 40.000 20.002 20.889 10.282 70.000 20.000 10.252 70.413 60.111 70.074 60.240 60.893 50.266 70.144 80.293 70.281 20.604 70.000 10.000 60.379 100.963 40.250 90.000 10.160 30.420 60.000 10.343 70.207 60.079 100.315 60.052 7
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
Mask3D Scannet2000.278 50.383 50.263 60.168 50.506 50.068 30.083 100.000 30.000 50.000 10.023 60.149 90.302 40.778 70.647 50.569 20.500 40.031 40.014 60.027 70.173 30.311 20.195 50.351 70.258 40.000 50.082 60.000 10.003 20.037 70.391 41.000 10.000 10.014 70.000 30.572 50.573 20.661 50.000 40.003 60.005 90.082 90.349 30.028 20.000 10.605 50.515 80.509 10.711 31.000 10.665 60.015 70.107 60.402 80.201 10.083 60.304 50.759 20.491 40.378 40.572 50.119 50.277 50.013 100.089 40.283 60.411 60.267 10.006 80.156 40.000 20.116 50.000 50.105 80.556 60.514 40.396 20.275 40.323 40.215 60.380 10.000 20.000 10.356 30.005 60.208 40.325 10.000 40.050 80.400 30.561 50.258 60.179 50.722 40.147 20.000 10.586 30.063 40.015 40.139 60.016 30.028 50.708 40.418 70.016 60.048 80.500 20.489 50.349 50.001 70.475 60.086 40.365 50.000 20.500 10.000 40.323 80.000 30.222 30.000 20.497 40.626 40.044 80.795 20.556 20.008 20.121 90.265 50.667 20.789 40.568 50.579 40.444 10.176 50.004 20.474 30.752 50.233 60.014 50.002 90.570 20.007 20.377 100.000 20.000 30.000 20.337 60.000 20.000 10.384 40.465 40.287 60.085 50.048 70.816 100.467 30.810 10.377 60.415 10.744 40.000 10.004 50.724 10.778 50.590 30.000 10.032 60.441 40.000 10.377 60.391 40.427 50.321 50.192 3
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
ACGP-ScanNet2000.381 10.486 10.362 10.275 10.597 30.133 20.333 10.500 10.394 10.000 10.008 80.483 20.512 31.000 10.649 40.497 40.792 30.032 30.556 20.179 30.170 40.307 40.291 30.480 20.114 61.000 10.242 20.000 10.037 10.512 30.365 51.000 10.000 10.396 30.000 30.608 30.184 50.643 60.009 20.007 40.271 10.209 10.304 50.000 40.000 10.731 10.678 20.248 50.779 11.000 10.647 80.080 10.288 30.423 70.000 40.396 20.435 30.903 10.499 30.400 10.676 20.247 10.329 30.500 10.062 70.462 20.673 10.144 20.574 10.252 30.000 20.365 40.000 50.336 10.733 40.556 10.412 10.312 20.581 20.524 20.313 30.000 20.000 10.349 40.037 30.301 10.036 30.194 20.143 60.600 10.652 30.677 20.314 10.772 20.000 40.000 10.444 70.104 10.031 30.486 10.077 10.472 11.000 10.635 30.500 20.454 10.500 20.782 20.449 40.018 40.538 30.069 50.406 30.002 10.146 70.014 30.795 20.000 30.139 60.001 10.686 10.815 10.541 10.753 40.556 20.007 30.284 10.330 40.778 10.926 10.792 40.785 10.444 10.380 20.000 30.514 20.821 20.346 20.197 20.065 50.494 40.000 50.395 90.000 20.000 30.000 20.391 40.000 20.000 10.546 20.543 10.548 20.438 30.240 50.895 40.388 50.569 40.694 20.197 60.824 10.000 10.060 20.612 61.000 10.832 20.000 10.461 10.752 10.000 10.486 20.850 10.466 20.400 40.112 6
Volt-SPFormerpermissive0.367 20.475 20.359 20.248 20.635 20.051 50.333 10.000 30.125 20.000 10.029 50.345 50.528 21.000 10.663 30.400 60.389 60.012 50.556 20.235 10.407 20.240 60.308 20.550 10.380 20.250 30.193 30.000 10.000 30.439 50.416 31.000 10.000 10.254 40.000 30.609 20.638 10.678 30.000 40.004 50.113 40.144 40.333 40.028 20.000 10.719 20.685 10.139 70.682 41.000 10.689 30.052 30.247 40.470 40.000 40.304 30.484 10.000 70.588 20.378 40.736 10.241 30.663 10.066 90.299 10.717 10.660 20.000 40.466 30.156 40.000 20.500 30.278 20.230 40.831 10.556 10.365 30.192 50.822 10.565 10.318 20.111 10.000 10.533 10.013 40.232 30.000 50.778 10.112 70.400 30.693 20.588 40.284 30.684 50.000 40.000 10.556 40.050 50.008 50.333 50.029 20.278 31.000 10.748 10.500 20.340 21.000 10.787 10.575 10.013 50.527 40.017 60.502 10.000 20.500 10.167 10.650 40.000 30.222 30.000 20.549 20.655 30.238 20.799 10.556 20.002 40.170 50.348 30.250 30.873 21.000 10.652 30.444 10.551 10.000 30.524 10.821 30.329 30.000 60.117 20.383 80.014 10.417 80.000 20.000 30.000 20.469 20.000 20.000 10.552 10.494 30.515 40.710 10.388 10.928 20.524 10.537 50.560 30.167 70.817 20.000 10.019 30.682 21.000 10.864 10.000 10.099 40.664 20.000 10.395 40.764 20.442 40.418 20.117 4
Kadir Yilmaz, Adrian Kruse, Tristan Höfer, Daan de Geus, Bastian Leibe: Volume Transformer: Revisiting Vanilla Transformers for 3D Scene Understanding.
CompetitorFormer-2000.328 40.439 30.303 40.223 40.543 40.044 60.333 10.044 20.000 50.000 10.099 10.444 30.296 50.850 50.722 20.820 10.444 50.047 10.083 50.183 20.562 10.243 50.312 10.380 40.192 51.000 10.143 50.000 10.000 30.484 40.259 61.000 10.000 10.500 10.000 30.650 10.221 40.771 10.004 30.010 30.043 60.120 60.366 20.054 10.000 10.689 30.641 40.500 20.663 51.000 10.673 40.049 40.400 10.479 30.014 30.267 40.455 20.083 60.400 50.400 10.663 30.243 20.464 20.192 60.076 60.427 30.620 30.025 30.013 70.322 20.000 20.677 20.333 10.178 50.808 20.556 10.356 40.345 10.119 70.346 40.312 40.000 20.000 10.305 50.116 20.137 60.000 50.065 30.171 50.314 60.575 40.487 50.303 20.820 10.000 40.000 10.655 10.088 20.373 10.430 30.011 40.103 40.835 30.569 50.125 40.123 50.500 20.774 30.504 30.019 30.465 70.353 20.475 20.000 20.500 10.000 40.712 30.050 20.667 10.000 20.396 50.555 50.120 50.786 30.069 50.000 50.182 40.390 20.000 80.831 31.000 10.679 20.111 40.110 60.000 30.450 50.868 10.277 40.083 30.069 30.471 50.001 40.428 70.000 20.000 30.000 20.421 30.043 10.000 10.358 50.456 50.518 30.237 40.256 30.945 10.271 60.632 30.534 50.208 30.730 50.000 10.140 10.658 31.000 10.452 60.000 10.082 50.441 50.000 10.472 30.060 80.454 30.469 10.384 1
DINO3D-Scannet200copyleft0.346 30.437 40.353 30.229 30.687 10.174 10.333 10.000 30.042 40.000 10.094 20.384 40.618 10.940 30.764 10.292 100.889 20.042 20.000 70.142 40.000 50.456 10.263 40.371 60.407 10.250 30.257 10.000 10.000 30.642 10.431 11.000 10.000 10.250 50.028 20.594 40.436 30.729 20.000 40.138 10.192 30.206 20.083 70.000 40.000 10.611 40.574 50.306 30.719 21.000 10.733 10.066 20.361 20.545 10.000 40.585 10.388 40.558 30.639 10.400 10.659 40.183 40.297 40.246 50.199 20.373 40.446 40.000 40.378 50.156 40.500 10.772 10.111 40.253 30.752 30.477 50.325 50.282 30.551 30.504 30.241 50.000 20.000 10.156 70.238 10.251 20.000 50.000 40.000 90.599 20.712 10.750 10.266 40.766 30.000 40.000 10.628 20.082 30.001 70.417 40.000 50.014 60.708 40.536 60.516 10.328 40.500 20.669 40.529 20.027 20.732 10.764 10.365 40.000 20.250 40.000 40.921 10.063 10.222 30.000 20.520 30.769 20.045 70.714 50.000 70.000 50.264 20.417 10.049 60.731 70.514 60.545 50.000 50.264 30.000 30.462 40.803 40.247 50.303 10.049 60.514 30.000 50.558 20.000 20.111 10.000 20.556 10.000 20.000 10.406 30.536 20.681 10.484 20.346 20.925 30.470 20.664 20.726 10.130 80.780 30.000 10.009 40.618 40.764 80.487 50.000 10.442 20.245 90.000 10.593 10.655 30.345 60.411 30.279 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
ODIN - Ins200permissive0.265 60.349 60.268 50.163 60.360 60.054 40.278 50.000 30.125 20.000 10.031 30.506 10.266 60.630 80.609 60.481 50.903 10.000 71.000 10.032 60.000 50.022 90.138 60.314 90.310 30.000 50.178 40.000 10.000 30.552 20.421 20.889 60.000 10.451 20.097 10.357 70.054 70.485 100.052 10.040 20.210 20.160 30.370 10.000 40.000 10.191 100.529 60.250 40.617 81.000 10.492 100.016 60.197 50.324 90.000 40.250 50.265 70.167 50.317 60.200 70.549 60.107 60.231 70.119 80.141 30.253 70.267 70.000 40.565 20.111 70.000 20.000 70.278 20.285 20.665 50.389 60.306 60.077 60.037 100.186 100.156 70.000 20.000 10.478 20.000 70.091 70.204 20.000 40.345 20.200 70.550 60.674 30.160 60.526 70.438 10.000 10.476 60.035 70.003 60.444 20.000 50.333 20.361 80.606 40.083 50.332 30.417 70.327 60.297 60.035 10.615 20.281 30.083 90.000 20.250 40.000 40.610 50.000 30.333 20.000 20.238 100.481 60.218 30.440 91.000 10.000 50.229 30.257 60.000 80.746 50.361 100.188 70.000 50.221 40.000 30.320 60.655 70.193 70.000 60.067 40.389 70.000 50.594 10.037 10.000 30.000 20.371 50.000 20.000 10.344 60.366 80.506 50.074 60.250 40.848 80.451 40.389 60.546 40.205 40.698 60.000 10.000 60.494 80.769 70.493 40.000 10.000 70.463 30.000 10.333 80.333 50.640 10.251 70.115 5
Minkowski 34D Inst.permissive0.130 90.246 90.083 90.043 100.299 90.000 100.278 50.000 30.000 50.000 10.022 70.175 80.122 70.537 90.521 70.400 60.000 70.000 70.000 70.008 80.000 50.048 80.076 80.182 100.000 90.000 50.022 90.000 10.000 30.000 80.141 100.000 70.000 10.000 80.000 30.210 90.063 60.547 90.000 40.000 70.000 100.100 70.026 100.000 40.000 10.241 80.488 90.000 90.564 101.000 10.672 50.000 80.021 90.486 20.000 40.000 80.067 90.000 70.194 100.033 90.415 90.026 90.025 100.271 20.004 90.094 100.142 100.000 40.000 90.111 70.000 20.000 70.000 50.088 90.083 100.278 70.110 90.000 90.082 90.199 90.137 80.000 20.000 10.000 80.000 70.041 90.000 50.000 40.308 30.067 80.280 80.016 90.101 80.373 100.000 40.000 10.319 90.007 90.000 80.000 80.000 50.000 80.028 100.355 100.000 70.101 60.444 60.289 70.114 100.000 80.394 80.000 70.032 100.000 20.000 80.000 40.201 100.000 30.000 70.000 20.384 60.248 90.000 100.529 80.000 70.000 50.133 80.020 100.089 50.720 80.500 80.099 90.000 50.000 100.000 30.238 90.334 100.190 80.000 60.000 100.317 100.000 50.472 30.000 20.000 30.000 20.094 100.000 20.000 10.082 100.236 90.004 100.019 90.000 80.883 60.061 100.262 70.217 90.000 90.557 100.000 10.000 60.460 90.761 90.156 100.000 10.000 70.259 80.000 10.394 50.019 90.084 90.232 90.000 10
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019
CSC-Pretrain Inst.permissive0.123 100.223 100.082 100.046 90.308 80.004 80.278 50.000 30.000 50.000 10.000 100.032 100.105 80.537 90.348 100.378 80.000 70.000 70.000 70.000 100.000 50.000 100.037 100.323 80.000 90.000 50.013 100.000 10.000 30.000 80.235 70.000 70.000 10.000 80.000 30.231 80.045 80.564 80.000 40.000 70.006 80.078 100.065 80.000 40.000 10.259 70.516 70.000 90.600 91.000 10.578 90.000 80.000 100.184 100.000 40.000 80.034 100.000 70.211 90.089 80.394 100.018 100.064 90.171 70.001 100.144 80.172 90.000 40.000 90.044 90.000 20.000 70.000 50.064 100.126 90.278 70.093 100.000 90.094 80.214 70.011 100.000 20.000 10.000 80.000 70.022 100.000 50.000 40.275 40.000 90.275 90.000 100.098 90.407 90.000 40.000 10.250 100.007 100.000 80.000 80.000 50.000 80.333 90.376 90.000 70.000 100.042 100.285 80.119 90.000 80.224 100.000 70.184 70.000 20.000 80.000 40.244 90.000 30.000 70.000 20.377 70.378 70.051 60.424 100.000 70.000 50.116 100.030 90.125 40.441 90.444 90.063 100.000 50.042 80.000 30.297 70.483 80.096 100.000 60.028 70.338 90.000 50.444 50.000 20.000 30.000 20.189 90.000 20.000 10.141 90.152 100.017 90.000 100.000 80.838 90.193 80.111 100.105 100.198 50.588 80.000 10.000 60.542 70.343 100.267 80.000 10.000 70.108 100.000 10.333 80.000 100.228 70.202 100.022 9
Ji Hou, Benjamin Graham, Matthias Nießner, Saining Xie: Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts. CVPR 2021
LGround Inst.permissive0.154 80.275 80.108 80.060 80.295 100.002 90.278 50.000 30.000 50.000 10.006 90.272 70.064 100.815 60.503 80.333 90.000 70.000 70.556 20.001 90.000 50.148 70.078 70.448 30.007 80.000 50.024 80.000 10.000 30.000 80.190 90.000 70.000 10.000 80.000 30.209 100.031 100.573 70.000 40.000 70.041 70.099 80.037 90.000 40.000 10.327 60.364 100.181 60.642 61.000 10.654 70.000 80.023 80.429 60.000 40.000 80.097 80.000 70.278 70.267 60.434 80.048 70.092 80.257 30.030 80.097 90.189 80.000 40.089 60.000 100.000 20.000 70.000 50.115 70.166 80.222 100.222 80.003 80.127 60.213 80.169 60.000 20.000 10.000 80.000 70.044 80.000 50.000 40.000 90.000 90.268 100.222 70.130 70.494 80.000 40.000 10.363 80.015 80.000 80.000 80.000 50.000 80.611 60.400 80.000 70.056 70.278 80.242 90.180 80.000 80.383 90.000 70.209 60.000 20.000 80.000 40.364 70.000 30.000 70.000 20.323 80.302 80.019 90.654 60.000 70.000 50.141 70.045 80.000 80.427 100.514 60.143 80.000 50.028 90.000 30.252 80.402 90.156 90.000 60.028 70.470 60.000 50.444 50.000 20.000 30.000 20.205 80.000 20.000 10.203 80.381 70.026 80.037 80.000 80.881 70.099 90.135 90.239 80.000 90.585 90.000 10.000 60.616 50.778 50.322 70.000 10.000 70.407 70.000 10.333 80.148 70.177 80.242 80.028 8
David Rozenberszki, Or Litany, Angela Dai: Language-Grounded Indoor 3D Semantic Segmentation in the Wild.