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 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 bysort bysort bysort bysort bysort bysort by
TD3D Scannet200permissive0.211 20.332 20.177 20.103 20.662 10.413 20.463 20.705 10.192 30.145 10.266 20.215 10.452 40.209 20.222 50.219 50.315 20.893 10.380 20.617 20.439 20.047 40.646 10.080 20.610 30.253 10.237 20.293 20.135 10.379 50.494 20.048 10.252 20.451 20.184 20.483 10.395 20.852 10.083 20.551 20.278 20.036 20.337 20.266 20.544 10.963 10.079 50.039 10.740 20.604 20.000 20.586 10.283 20.282 20.059 20.633 30.028 20.004 20.559 20.309 20.420 20.028 51.000 10.000 10.456 10.411 10.372 10.060 40.046 40.000 20.040 40.694 10.083 20.000 20.000 10.000 20.000 30.083 40.252 20.260 50.200 10.160 10.669 20.111 20.000 20.000 10.006 20.169 20.000 10.007 10.296 20.032 10.074 10.139 30.000 20.321 20.031 10.108 20.088 20.157 10.000 10.231 50.026 50.000 20.000 10.356 20.052 20.000 10.240 10.147 10.000 10.015 20.046 30.144 30.073 30.414 10.222 40.000 10.806 10.343 30.486 30.000 10.008 10.038 20.083 10.002 10.028 20.074 20.032 20.150 20.039 20.008 10.000 10.250 40.000 10.125 40.000 10.052 20.260 30.000 10.143 50.000 10.000 10.543 20.207 20.404 10.000 10.003 20.000 10.000 20.000 10.037 20.093 40.272 20.342 10.039 40.281 20.249 30.224 10.000 20.000 10.074 10.000 10.000 10.000 20.278 20.000 10.000 20.889 10.323 10.000 20.014 10.000 20.000 1
Maksim Kolodiazhnyi, Anna Vorontsova, Anton Konushin, Danila Rukhovich: Top-Down Beats Bottom-Up in 3D Instance Segmentation. WACV 2024
Minkowski 34D Inst.permissive0.130 40.246 40.083 40.043 50.547 50.236 40.415 40.672 20.141 50.133 30.067 40.000 30.521 20.114 50.238 40.289 20.232 40.883 20.182 50.373 50.486 10.076 30.488 40.022 40.529 40.199 50.110 40.217 40.100 20.460 40.319 40.000 30.025 50.472 10.000 30.394 30.210 40.537 40.004 40.000 30.083 50.000 50.299 40.061 50.201 50.761 40.084 40.008 30.720 30.557 50.000 20.317 50.280 30.094 50.020 50.564 50.000 40.000 30.400 30.048 40.259 40.101 31.000 10.000 10.190 30.142 50.094 50.137 30.089 30.000 20.101 10.355 50.000 30.000 20.000 10.000 20.000 30.444 20.082 50.384 20.000 50.000 30.334 50.004 50.000 20.000 10.000 30.041 40.000 10.000 30.026 50.000 30.000 30.000 40.000 20.082 50.022 30.000 50.021 40.088 40.000 10.241 40.033 40.000 20.000 10.067 30.000 50.000 10.000 30.000 30.000 10.000 40.026 40.262 20.016 40.000 40.278 10.000 10.500 40.394 10.028 50.000 10.000 30.000 30.000 30.000 20.000 30.019 40.000 30.000 30.000 30.000 30.000 10.156 50.000 10.032 50.000 10.000 30.194 50.000 10.248 40.000 10.000 10.099 40.019 40.308 20.000 10.000 30.000 10.000 20.000 10.007 40.122 20.000 30.175 30.063 20.000 40.271 10.000 50.000 20.000 10.000 50.000 10.000 10.000 20.278 20.000 10.000 20.000 20.111 30.000 20.000 20.000 20.000 1
C. Choy, J. Gwak, S. Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019
LGround Inst.permissive0.154 30.275 30.108 30.060 30.573 30.381 30.434 30.654 40.190 40.141 20.097 30.000 30.503 30.180 30.252 30.242 40.242 30.881 30.448 10.494 30.429 30.078 20.364 50.024 30.654 20.213 40.222 30.239 30.099 30.616 20.363 30.000 30.092 30.444 30.000 30.383 40.209 50.815 20.030 30.000 30.166 30.002 40.295 50.099 40.364 20.778 20.177 30.001 40.427 50.585 40.000 20.470 30.268 50.205 30.045 30.642 20.007 30.000 30.333 50.148 30.407 30.130 21.000 10.000 10.156 40.189 30.097 40.169 20.000 50.000 20.056 20.400 30.000 30.000 20.000 10.000 20.556 10.278 30.203 30.323 40.019 40.000 30.402 40.026 30.000 20.000 10.000 30.044 30.000 10.000 30.037 40.000 30.000 30.181 20.000 20.127 30.006 40.028 40.023 30.115 20.000 10.327 20.267 20.000 20.000 10.000 40.028 30.000 10.000 30.000 30.000 10.003 30.048 20.135 40.222 20.089 20.278 10.000 10.514 30.333 40.611 20.000 10.000 30.000 30.000 30.000 20.000 30.037 30.000 30.000 30.000 30.000 30.000 10.322 20.000 10.209 20.000 10.000 30.278 20.000 10.302 30.000 10.000 10.143 30.148 30.000 50.000 10.000 30.000 10.000 20.000 10.015 30.064 50.000 30.272 20.031 50.000 40.257 20.028 20.000 20.000 10.041 20.000 10.000 10.000 20.222 50.000 10.000 20.000 20.000 50.000 20.000 20.000 20.000 1
David Rozenberszki, Or Litany, Angela Dai: Language-Grounded Indoor 3D Semantic Segmentation in the Wild.
Mask3D Scannet2000.278 10.383 10.263 10.168 10.661 20.465 10.572 10.665 30.391 10.121 40.304 10.015 20.647 10.349 10.474 10.489 10.321 10.816 50.351 30.722 10.402 40.195 10.515 30.082 10.795 10.215 20.396 10.377 10.082 40.724 10.586 10.015 20.277 10.377 50.201 10.475 20.572 10.778 30.089 10.759 10.556 10.068 10.506 10.467 10.323 30.778 20.427 10.027 20.789 10.744 10.003 10.570 20.561 10.337 10.265 10.711 10.258 10.031 10.569 10.311 10.441 10.179 11.000 10.000 10.233 20.411 20.283 20.380 10.667 10.016 10.048 30.418 20.139 10.173 10.000 10.086 10.014 20.500 10.384 10.497 10.044 30.032 20.752 10.287 10.003 10.000 10.007 10.208 10.000 10.001 20.349 10.008 20.014 20.509 10.500 10.323 10.023 20.176 10.107 10.105 30.000 10.605 10.378 10.016 10.000 10.400 10.192 10.000 10.048 20.037 20.000 10.275 10.119 10.810 10.258 10.006 30.083 50.000 10.568 20.377 20.708 10.000 10.005 20.147 10.014 20.000 20.556 10.085 10.325 10.500 10.083 10.004 20.000 10.590 10.000 10.365 10.000 10.116 10.491 10.000 10.626 10.000 10.000 10.579 10.391 10.050 40.000 10.028 10.000 10.222 10.000 10.063 10.302 10.356 10.149 40.573 10.415 10.013 50.002 40.004 10.000 10.005 40.000 10.000 10.444 10.514 10.000 10.028 10.000 20.156 20.267 10.000 21.000 10.000 1
Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, Bastian Leibe: Mask3D for 3D Semantic Instance Segmentation. ICRA 2023
CSC-Pretrain Inst.permissive0.123 50.223 50.082 50.046 40.564 40.152 50.394 50.578 50.235 20.116 50.034 50.000 30.348 50.119 40.297 20.285 30.202 50.838 40.323 40.407 40.184 50.037 50.516 20.013 50.424 50.214 30.093 50.105 50.078 50.542 30.250 50.000 30.064 40.444 30.000 30.224 50.231 30.537 40.001 50.000 30.126 40.004 30.308 30.193 30.244 40.343 50.228 20.000 50.441 40.588 30.000 20.338 40.275 40.189 40.030 40.600 40.000 40.000 30.378 40.000 50.108 50.098 41.000 10.000 10.096 50.172 40.144 30.011 50.125 20.000 20.000 50.376 40.000 30.000 20.000 10.000 20.000 30.042 50.141 40.377 30.051 20.000 30.483 30.017 40.000 20.000 10.000 30.022 50.000 10.000 30.065 30.000 30.000 30.000 40.000 20.094 40.000 50.042 30.000 50.064 50.000 10.259 30.089 30.000 20.000 10.000 40.022 40.000 10.000 30.000 30.000 10.000 40.018 50.111 50.000 50.000 40.278 10.000 10.444 50.333 40.333 40.000 10.000 30.000 30.000 30.000 20.000 30.000 50.000 30.000 30.000 30.000 30.000 10.267 30.000 10.184 30.000 10.000 30.211 40.000 10.378 20.000 10.000 10.063 50.000 50.275 30.000 10.000 30.000 10.000 20.000 10.007 50.105 30.000 30.032 50.045 30.198 30.171 40.028 20.000 20.000 10.006 30.000 10.000 10.000 20.278 20.000 10.000 20.000 20.044 40.000 20.000 20.000 20.000 1
Ji Hou, Benjamin Graham, Matthias Nießner, Saining Xie: Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts. CVPR 2021