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 apchairtabledoorcouchcabinetshelfdeskoffice 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 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 by
AQ3D-ScanNet2000.385 10.522 10.374 10.234 30.759 20.510 30.671 30.754 10.469 10.221 40.507 10.303 20.687 30.578 10.462 40.777 30.455 20.923 40.492 20.771 30.670 10.358 10.683 20.209 30.799 20.554 20.435 10.590 30.203 30.704 20.655 10.030 60.520 20.455 40.150 30.681 20.641 21.000 10.077 60.020 70.748 40.127 30.698 10.448 50.667 41.000 10.406 60.239 10.928 10.825 10.005 50.773 10.585 40.398 40.422 10.676 50.147 60.013 50.527 40.458 10.618 30.231 51.000 10.574 10.331 30.733 10.451 30.363 21.000 10.066 20.267 50.801 10.556 10.148 50.132 10.000 70.533 51.000 10.532 30.501 40.475 20.664 10.792 50.617 20.009 20.009 30.000 50.341 10.000 20.028 20.296 60.029 20.158 30.139 70.889 20.598 20.023 60.454 20.306 30.276 30.000 20.399 60.570 10.083 50.438 20.400 30.551 10.500 10.367 20.548 30.337 10.304 30.224 40.657 30.658 40.292 60.500 10.000 21.000 10.333 81.000 10.000 10.075 30.000 40.451 20.000 30.556 20.395 40.000 50.140 80.025 80.000 30.000 30.864 10.000 10.479 20.000 10.424 40.565 30.000 10.856 10.000 10.000 10.482 70.667 30.125 70.167 50.667 10.000 10.444 20.000 10.052 50.549 20.468 30.383 50.073 60.410 20.333 20.101 30.083 20.000 10.299 10.000 10.000 20.444 10.444 60.000 30.037 20.000 20.156 40.141 30.000 40.167 70.000 2
DINO3D-Scannet200copyleft0.346 40.437 50.353 40.229 40.729 30.536 20.659 50.733 20.431 20.264 20.388 50.001 80.764 10.529 30.462 50.669 50.411 40.925 30.371 70.766 40.545 20.263 50.574 60.257 10.714 60.504 40.325 60.726 10.206 20.618 50.628 30.066 20.297 50.558 20.000 50.732 10.594 50.940 40.199 20.558 30.752 30.174 10.687 20.470 20.921 10.764 90.345 70.142 50.731 80.780 40.138 10.514 40.712 10.556 10.417 20.719 20.407 10.042 20.292 110.456 20.245 100.266 41.000 10.042 50.247 60.446 50.373 50.241 60.049 70.000 60.328 40.536 70.417 50.000 60.000 30.764 10.000 80.500 30.406 40.520 30.045 80.442 30.803 40.681 10.000 40.000 50.000 50.251 30.000 20.027 30.083 80.000 60.303 10.306 30.889 20.551 40.094 20.264 40.361 20.253 40.000 30.611 40.400 20.516 10.000 50.599 20.279 30.000 40.346 30.642 10.111 50.282 40.183 50.664 20.750 10.378 50.333 20.500 10.514 70.593 10.708 50.000 10.238 10.000 40.250 60.111 10.000 80.484 20.000 50.250 40.585 10.000 30.063 10.487 60.000 10.365 50.000 10.772 10.639 10.000 10.769 30.000 10.000 10.545 50.655 40.000 100.250 30.014 70.000 10.222 40.000 10.082 30.618 10.156 80.384 40.436 30.130 90.246 60.049 70.009 50.000 10.192 40.000 10.000 20.000 60.477 50.028 20.000 50.000 20.156 40.000 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
Volt-SPFormerpermissive0.367 30.475 30.359 30.248 20.678 40.494 40.736 10.689 40.416 40.170 60.484 20.008 60.663 40.575 20.524 10.787 10.418 30.928 20.550 10.684 60.470 50.308 30.685 10.193 40.799 10.565 10.365 40.560 40.144 50.682 30.556 50.052 30.663 10.417 90.000 50.527 50.609 31.000 10.299 10.000 80.831 10.051 60.635 30.524 10.650 51.000 10.442 40.235 20.873 30.817 30.004 60.383 90.693 20.469 20.348 40.682 40.380 20.012 60.400 70.240 70.664 20.284 31.000 10.125 30.329 40.660 30.717 10.318 30.250 40.029 30.340 20.748 20.333 60.407 20.000 30.017 60.556 21.000 10.552 10.549 20.238 30.099 50.821 30.515 50.000 40.000 50.014 10.232 40.111 10.013 60.333 40.002 50.000 70.139 70.389 70.822 10.029 50.551 10.247 50.230 50.000 30.719 20.378 50.500 20.778 10.400 30.117 50.000 40.388 10.439 60.278 30.192 60.241 30.537 60.588 50.466 30.333 20.000 21.000 10.395 41.000 10.000 10.013 50.000 40.254 50.000 30.556 20.710 10.000 50.500 10.304 30.000 30.000 30.864 10.000 10.502 10.000 10.500 30.588 20.000 10.655 40.000 10.000 10.652 30.764 20.112 80.250 30.278 40.000 10.222 40.000 10.050 60.528 30.533 10.345 60.638 10.167 80.066 100.117 20.019 40.000 10.113 50.000 10.000 20.444 10.556 10.000 30.028 30.000 20.156 40.000 50.167 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-ScanNet2000.381 20.486 20.362 20.275 10.643 70.543 10.676 20.647 90.365 60.284 10.435 40.031 40.649 50.449 50.514 20.782 20.400 50.895 50.480 30.772 20.423 80.291 40.678 30.242 20.753 50.524 30.412 20.694 20.209 10.612 70.444 80.080 10.329 40.395 100.000 50.538 40.608 41.000 10.062 80.903 10.733 50.133 20.597 40.388 60.795 21.000 10.466 20.179 40.926 20.824 20.007 40.494 50.652 30.391 50.330 50.779 10.114 70.032 30.497 50.307 50.752 10.314 11.000 10.394 20.346 20.673 20.462 20.313 40.778 20.077 10.454 10.635 40.486 20.170 40.001 20.069 50.556 20.500 30.546 20.686 10.541 10.461 20.821 20.548 30.037 10.009 20.000 50.301 20.000 20.018 50.304 50.007 40.197 20.248 50.792 40.581 30.008 90.380 30.288 40.336 10.000 30.731 10.400 20.500 20.194 30.600 10.112 70.500 10.240 60.512 40.000 60.312 20.247 10.569 50.677 20.574 10.333 20.000 20.792 50.486 21.000 10.000 10.037 40.000 40.396 40.000 30.556 20.438 30.036 30.146 70.396 20.000 30.000 30.832 30.000 10.406 40.000 10.365 50.499 40.000 10.815 20.000 10.000 10.785 10.850 10.143 61.000 10.472 20.000 10.139 70.000 10.104 10.512 40.349 50.483 20.184 50.197 70.500 10.065 60.060 30.000 10.271 20.000 10.000 20.444 10.556 10.000 30.000 50.000 20.252 30.144 20.014 31.000 10.002 1
Rongkun Yang, Ye Zhang, Longguang Wang, Zhiheng Fu, Lian Xu, Yulan Guo: Beyond Context Bias: Adaptive Instance Placement for Robust 3D Instance Segmentation.
CompetitorFormer-2000.328 50.439 40.303 50.223 50.771 10.456 60.663 40.673 50.259 70.182 50.455 30.373 10.722 20.504 40.450 60.774 40.469 10.945 10.380 50.820 10.479 40.312 20.641 50.143 60.786 40.346 50.356 50.534 60.120 70.658 40.655 20.049 40.464 30.428 80.014 40.465 80.650 10.850 60.076 70.083 60.808 20.044 70.543 50.271 70.712 31.000 10.454 30.183 30.831 40.730 60.010 30.471 60.575 50.421 30.390 30.663 60.192 50.047 10.820 10.243 60.441 60.303 21.000 10.000 60.277 50.620 40.427 40.312 50.000 90.011 50.123 60.569 60.430 40.562 10.000 30.353 20.083 60.500 30.358 60.396 60.120 60.082 60.868 10.518 40.000 40.004 40.001 40.137 70.000 20.019 40.366 20.000 60.083 40.500 20.444 60.119 80.099 10.110 70.400 10.178 60.000 30.689 30.400 20.125 40.065 40.314 70.384 20.044 30.256 40.484 50.333 20.345 10.243 20.632 40.487 60.013 80.333 20.000 21.000 10.472 30.835 40.000 10.116 20.000 40.500 10.000 30.069 60.237 50.000 50.500 10.267 40.000 30.050 20.452 70.000 10.475 30.000 10.677 20.400 60.000 10.555 60.000 10.000 10.679 20.060 90.171 51.000 10.103 50.000 10.667 10.000 10.088 20.296 60.305 60.444 30.221 40.208 40.192 70.069 40.140 10.000 10.043 70.000 10.043 10.111 50.556 10.000 30.054 10.000 20.322 20.025 40.000 41.000 10.000 2
Mask3D Scannet2000.278 60.383 60.263 70.168 60.661 60.465 50.572 60.665 70.391 50.121 100.304 60.015 50.647 60.349 60.474 30.489 60.321 60.816 110.351 80.722 50.402 90.195 60.515 90.082 70.795 30.215 70.396 30.377 70.082 100.724 10.586 40.015 80.277 60.377 110.201 10.475 70.572 60.778 80.089 40.759 20.556 70.068 40.506 60.467 30.323 90.778 60.427 50.027 80.789 50.744 50.003 70.570 30.561 60.337 70.265 60.711 30.258 40.031 40.569 20.311 30.441 50.179 61.000 10.000 60.233 70.411 70.283 70.380 10.667 30.016 40.048 90.418 80.139 70.173 30.000 30.086 40.014 70.500 30.384 50.497 50.044 90.032 70.752 60.287 70.003 30.000 50.007 20.208 50.000 20.001 80.349 30.008 30.014 60.509 10.500 50.323 50.023 70.176 60.107 70.105 90.000 30.605 50.378 50.016 70.000 50.400 30.192 40.000 40.048 80.037 80.000 60.275 50.119 60.810 10.258 70.006 90.083 110.000 20.568 60.377 60.708 50.000 10.005 70.147 20.014 80.000 30.556 20.085 60.325 10.500 10.083 60.004 20.000 30.590 40.000 10.365 60.000 10.116 60.491 50.000 10.626 50.000 10.000 10.579 40.391 50.050 90.000 60.028 60.000 10.222 40.000 10.063 40.302 50.356 40.149 100.573 20.415 10.013 110.002 100.004 60.000 10.005 100.000 10.000 20.444 10.514 40.000 30.028 30.000 20.156 40.267 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
ODIN - Ins200permissive0.265 70.349 70.268 60.163 70.485 110.366 90.549 70.492 110.421 30.229 30.265 80.003 70.609 70.297 70.320 70.327 70.251 80.848 90.314 100.526 80.324 100.138 70.529 70.178 50.440 100.186 110.306 70.546 50.160 40.494 90.476 70.016 70.231 80.594 10.000 50.615 30.357 80.630 90.141 30.167 50.665 60.054 50.360 70.451 40.610 60.769 80.640 10.032 70.746 60.698 70.040 20.389 80.550 70.371 60.257 70.617 90.310 30.000 80.481 60.022 100.463 40.160 71.000 10.125 30.193 80.267 80.253 80.156 80.000 90.000 60.332 30.606 50.444 30.000 60.000 30.281 31.000 10.417 80.344 70.238 110.218 40.000 80.655 80.506 60.000 40.052 10.000 50.091 80.000 20.035 10.370 10.000 60.000 70.250 40.903 10.037 110.031 30.221 50.197 60.285 20.037 10.191 110.200 80.083 50.000 50.200 80.115 60.000 40.250 50.552 20.278 30.077 70.107 70.389 70.674 30.565 20.278 60.000 20.361 110.333 80.361 90.000 10.000 80.438 10.451 20.000 31.000 10.074 70.204 20.250 40.250 50.000 30.000 30.493 50.000 10.083 100.000 10.000 80.317 70.000 10.481 70.000 10.000 10.188 80.333 60.345 20.000 60.333 30.000 10.333 30.000 10.035 80.266 70.478 20.506 10.054 80.205 50.119 90.067 50.000 70.000 10.210 30.000 10.000 20.000 60.389 70.097 10.000 50.000 20.111 80.000 50.000 40.889 60.000 2
TD3D Scannet200permissive0.211 80.332 80.177 80.103 80.662 50.413 70.463 80.705 30.192 90.145 70.266 70.215 30.452 100.209 80.222 110.219 110.315 70.893 60.380 60.617 70.439 60.047 100.646 40.080 80.610 80.253 60.237 80.293 80.135 60.379 110.494 60.048 50.252 70.451 50.184 20.483 60.395 70.852 50.083 50.551 40.278 80.036 80.337 80.266 80.544 70.963 50.079 110.039 60.740 70.604 80.000 80.586 20.283 80.282 80.059 80.633 80.028 80.004 70.559 30.309 40.420 70.028 111.000 10.000 60.456 10.411 60.372 60.060 100.046 80.000 60.040 100.694 30.083 80.000 60.000 30.000 70.000 80.083 100.252 80.260 100.200 50.160 40.669 70.111 80.000 40.000 50.006 30.169 60.000 20.007 70.296 60.032 10.074 50.139 70.000 80.321 60.031 40.108 80.088 80.157 70.000 30.231 100.026 110.000 80.000 50.356 60.052 80.000 40.240 70.147 70.000 60.015 80.046 90.144 90.073 90.414 40.222 100.000 20.806 40.343 70.486 80.000 10.008 60.038 30.083 70.002 20.028 70.074 70.032 40.150 60.039 70.008 10.000 30.250 100.000 10.125 90.000 10.052 70.260 90.000 10.143 110.000 10.000 10.543 60.207 70.404 10.000 60.003 80.000 10.000 80.000 10.037 70.093 100.272 70.342 70.039 100.281 30.249 50.224 10.000 70.000 10.074 60.000 10.000 20.000 60.278 80.000 30.000 50.889 10.323 10.000 50.014 20.000 80.000 2
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
CSC-Pretrain Inst.permissive0.123 110.223 110.082 110.046 100.564 90.152 110.394 110.578 100.235 80.116 110.034 110.000 90.348 110.119 100.297 80.285 90.202 110.838 100.323 90.407 100.184 110.037 110.516 80.013 110.424 110.214 80.093 110.105 110.078 110.542 80.250 110.000 90.064 100.444 60.000 50.224 110.231 90.537 100.001 110.000 80.126 100.004 90.308 90.193 90.244 100.343 110.228 80.000 110.441 100.588 90.000 80.338 100.275 100.189 100.030 100.600 100.000 100.000 80.378 90.000 110.108 110.098 101.000 10.000 60.096 110.172 100.144 90.011 110.125 50.000 60.000 110.376 100.000 90.000 60.000 30.000 70.000 80.042 110.141 100.377 80.051 70.000 80.483 90.017 100.000 40.000 50.000 50.022 110.000 20.000 90.065 90.000 60.000 70.000 100.000 80.094 90.000 110.042 90.000 110.064 110.000 30.259 80.089 90.000 80.000 50.000 100.022 100.000 40.000 90.000 90.000 60.000 100.018 110.111 110.000 110.000 100.278 60.000 20.444 100.333 80.333 100.000 10.000 80.000 40.000 90.000 30.000 80.000 110.000 50.000 90.000 90.000 30.000 30.267 90.000 10.184 80.000 10.000 80.211 100.000 10.378 80.000 10.000 10.063 110.000 110.275 40.000 60.000 90.000 10.000 80.000 10.007 110.105 90.000 90.032 110.045 90.198 60.171 80.028 80.000 70.000 10.006 90.000 10.000 20.000 60.278 80.000 30.000 50.000 20.044 100.000 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
Minkowski 34D Inst.permissive0.130 100.246 100.083 100.043 110.547 100.236 100.415 100.672 60.141 110.133 90.067 100.000 90.521 80.114 110.238 100.289 80.232 100.883 70.182 110.373 110.486 30.076 90.488 100.022 100.529 90.199 100.110 100.217 100.100 80.460 100.319 100.000 90.025 110.472 30.000 50.394 90.210 100.537 100.004 100.000 80.083 110.000 110.299 100.061 110.201 110.761 100.084 100.008 90.720 90.557 110.000 80.317 110.280 90.094 110.020 110.564 110.000 100.000 80.400 70.048 90.259 90.101 91.000 10.000 60.190 90.142 110.094 110.137 90.089 60.000 60.101 70.355 110.000 90.000 60.000 30.000 70.000 80.444 70.082 110.384 70.000 110.000 80.334 110.004 110.000 40.000 50.000 50.041 100.000 20.000 90.026 110.000 60.000 70.000 100.000 80.082 100.022 80.000 110.021 100.088 100.000 30.241 90.033 100.000 80.000 50.067 90.000 110.000 40.000 90.000 90.000 60.000 100.026 100.262 80.016 100.000 100.278 60.000 20.500 90.394 50.028 110.000 10.000 80.000 40.000 90.000 30.000 80.019 100.000 50.000 90.000 90.000 30.000 30.156 110.000 10.032 110.000 10.000 80.194 110.000 10.248 100.000 10.000 10.099 100.019 100.308 30.000 60.000 90.000 10.000 80.000 10.007 100.122 80.000 90.175 90.063 70.000 100.271 30.000 110.000 70.000 10.000 110.000 10.000 20.000 60.278 80.000 30.000 50.000 20.111 80.000 50.000 40.000 80.000 2
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019
LGround Inst.permissive0.154 90.275 90.108 90.060 90.573 80.381 80.434 90.654 80.190 100.141 80.097 90.000 90.503 90.180 90.252 90.242 100.242 90.881 80.448 40.494 90.429 70.078 80.364 110.024 90.654 70.213 90.222 90.239 90.099 90.616 60.363 90.000 90.092 90.444 60.000 50.383 100.209 110.815 70.030 90.000 80.166 90.002 100.295 110.099 100.364 80.778 60.177 90.001 100.427 110.585 100.000 80.470 70.268 110.205 90.045 90.642 70.007 90.000 80.333 100.148 80.407 80.130 81.000 10.000 60.156 100.189 90.097 100.169 70.000 90.000 60.056 80.400 90.000 90.000 60.000 30.000 70.556 20.278 90.203 90.323 90.019 100.000 80.402 100.026 90.000 40.000 50.000 50.044 90.000 20.000 90.037 100.000 60.000 70.181 60.000 80.127 70.006 100.028 100.023 90.115 80.000 30.327 70.267 70.000 80.000 50.000 100.028 90.000 40.000 90.000 90.000 60.003 90.048 80.135 100.222 80.089 70.278 60.000 20.514 70.333 80.611 70.000 10.000 80.000 40.000 90.000 30.000 80.037 90.000 50.000 90.000 90.000 30.000 30.322 80.000 10.209 70.000 10.000 80.278 80.000 10.302 90.000 10.000 10.143 90.148 80.000 100.000 60.000 90.000 10.000 80.000 10.015 90.064 110.000 90.272 80.031 110.000 100.257 40.028 80.000 70.000 10.041 80.000 10.000 20.000 60.222 110.000 30.000 50.000 20.000 110.000 50.000 40.000 80.000 2
David Rozenberszki, Or Litany, Angela Dai: Language-Grounded Indoor 3D Semantic Segmentation in the Wild.