Semantic Scene Completion on NYU (test)
0Ceiling ErrorLin et al.
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Lin et al.Resolution=(240, 60), Trained on=NYU, RGB-D based=false2020.03 | 0 | 11.7 | 13.3 | 14.1 | 9.4 | 29 | 24 | 6 | 7 | 16.2 | 1.1 | 12 | — | |
| Lin et al.training_data_constraint=NYU only2019.09 | 0 | 11.7 | 13.3 | 14.1 | 9.4 | 29 | 24 | 6 | 7 | 16.2 | 1.1 | 12 | — | |
| TS3DResolution=(240, 60), Trained on=NYU, RGB-D based=true2020.03 | 9.7 | 93.4 | 25.5 | 21 | 17.4 | 55.9 | 49.2 | 17 | 27.5 | 39.4 | 19.3 | 34.1 | — | |
| TS3DInput=RGB+D, Resolution=(240, 60)2021.12 | 9.7 | 93.4 | 25.5 | 21 | 17.4 | 55.9 | 49.2 | 17 | 27.5 | 39.4 | 19.3 | — | 0.341 | |
| Geiger et al.Resolution=(240, 60), Trained on=NYU, RGB-D based=false2020.03 | 10.2 | 62.5 | 19.1 | 5.8 | 8.5 | 40.6 | 27.7 | 7 | 6 | 22.6 | 5.9 | 19.6 | — | |
| Geiger and Wangtraining_data_constraint=NYU only2019.09 | 10.2 | 62.5 | 19.1 | 5.8 | 8.5 | 40.6 | 27.7 | 7 | 6 | 22.6 | 5.9 | 19.6 | — | |
| SSCNetResolution=(240, 60), Trained on=NYU, RGB-D based=false2020.03 | 15.1 | 94.7 | 24.4 | 0 | 12.6 | 32.1 | 35 | 13 | 7.8 | 27.1 | 10.1 | 24.7 | — | |
| SSCNet2019.09 | 15.1 | 94.6 | 24.7 | 10.8 | 17.3 | 53.2 | 45.9 | 15.9 | 13.9 | 31.1 | 12.6 | 30.5 | — | |
| SSCNetInput=D, Resolution=(240, 60)2021.12 | 15.1 | 94.7 | 24.4 | 0 | 12.6 | 32.1 | 35 | 13 | 7.8 | 27.1 | 10.1 | — | 0.247 | |
| SSCNetInputs=TSDF, Input depth condition=noisy depth2023.03 | 15.1 | 94.7 | 24.4 | 0 | 12.6 | 32.1 | 35 | 13 | 7.8 | 27.1 | 10.1 | 24.7 | — | |
| SATNet-TNetFuseResolution=(60, 60), Trained on=NYU+SUNCG, RGB-D based=true2020.03 | 17.3 | 92.1 | 28 | 16.6 | 19.3 | 57.5 | 53.8 | 17.2 | 18.5 | 38.4 | 18.9 | 34.4 | — | |
| SaTNet2019.09 | 17.3 | 92.1 | 28 | 16.6 | 19.3 | 57.5 | 53.8 | 17.7 | 18.5 | 38.4 | 18.9 | 34.4 | — | |
| SATNetInput=RGB+D, Resolution=(60, 60)2021.12 | 17.3 | 92.1 | 28 | 16.6 | 19.3 | 57.5 | 53.8 | 17.2 | 18.5 | 38.4 | 18.9 | — | 0.344 | |
| ESSCNetResolution=(240, 60), Trained on=NYU, RGB-D based=false2020.03 | 17.5 | 75.4 | 25.8 | 6.7 | 15.3 | 53.8 | 42.4 | 11.2 | 0 | 33.4 | 11.8 | 26.7 | — | |
| ESSCNetInput=D, Resolution=(240, 60)2021.12 | 17.5 | 75.4 | 25.8 | 6.7 | 15.3 | 53.8 | 42.4 | 11.2 | 0 | 33.4 | 11.8 | — | 0.267 | |
| VVNetR-120Resolution=(120, 60), Trained on=NYU+SUNCG, RGB-D based=false2020.03 | 19.3 | 94.8 | 28 | 12.2 | 19.6 | 57 | 50.5 | 17.6 | 11.9 | 35.6 | 15.3 | 32.9 | — | |
| VVNet2019.09 | 19.3 | 94.8 | 28 | 12.2 | 19.6 | 57 | 50.5 | 17.6 | 11.9 | 35.6 | 15.3 | 32.9 | — | |
| VVNetInput=D, Resolution=(120, 60)2021.12 | 19.3 | 94.8 | 28 | 12.2 | 19.6 | 57 | 50.5 | 17.6 | 11.9 | 35.6 | 15.3 | — | 0.329 | |
| DDRNetResolution=(240, 60), Trained on=NYU, RGB-D based=true2020.03 | 21.1 | 92.2 | 33.5 | 6.8 | 14.8 | 48.3 | 42.3 | 13.2 | 13.9 | 35.3 | 13.2 | 30.4 | — | |
| DDRNetInput=RGB+D, Resolution=(60, 60)2021.12 | 21.1 | 92.2 | 33.5 | 6.8 | 14.8 | 48.3 | 42.3 | 13.2 | 13.9 | 35.3 | 13.2 | — | 0.304 | |
| DDRNetInputs=RGB+D, Input depth condition=noisy depth2023.03 | 21.1 | 92.2 | 33.5 | 6.8 | 14.8 | 48.3 | 42.3 | 13.2 | 13.9 | 35.3 | 13.2 | 30.4 | — | |
| AIC-NetInput=RGB+D, Resolution=(60, 60)2021.12 | 23.2 | 90.8 | 32.3 | 14.8 | 18.2 | 51.1 | 44.8 | 15.2 | 22.4 | 38.3 | 15.7 | — | 0.333 | |
| AIC-NetInputs=RGB+D, Input depth condition=noisy depth2023.03 | 23.2 | 90.8 | 32.3 | 14.8 | 18.2 | 51.1 | 44.8 | 15.2 | 22.4 | 38.3 | 15.7 | 33.3 | — | |
| CCPNetResolution=(240, 240), Trained on=NYU, RGB-D based=false2020.03 | 23.5 | 96.3 | 35.7 | 20.2 | 25.8 | 61.4 | 56.1 | 18.1 | 28.1 | 37.8 | 20.1 | 38.5 | — | |
| CCPNetInput=D, Resolution=(240, 240)2021.12 | 23.5 | 96.3 | 35.7 | 20.2 | 25.8 | 61.4 | 56.1 | 18.1 | 28.1 | 37.8 | 20.1 | — | 0.385 | |
| CCPNetInputs=TSDF, Input depth condition=noisy depth2023.03 | 23.5 | 96.3 | 35.7 | 20.2 | 25.8 | 61.4 | 56.1 | 18.1 | 28.1 | 37.8 | 20.1 | 38.5 | — | |
| IPF-SPCNetInputs=RGB+Point, Input depth condition=noisy depth2023.03 | 32.7 | 66 | 41.2 | 17.2 | 34.7 | 55.3 | 47 | 21.7 | 12.5 | 38.4 | 19.2 | 35.1 | — | |
| SPCNetInput=D, Resolution=(240, 60)2021.12 | 33.8 | 64.4 | 38.3 | 7.5 | 30.7 | 53.4 | 42.6 | 19.7 | 5.5 | 34.2 | 13.9 | — | 0.313 | |
| 3D-RecGAN2019.09 | 35.3 | 70.3 | 24.1 | 3.8 | 11.9 | 47.4 | 43.1 | 11.4 | 16.9 | 30.6 | 7.2 | 27.5 | — | |
| ForkNetablation=without completion branch2019.09 | 35.8 | 94.1 | 28.9 | 19.2 | 16.8 | 61.4 | 53.5 | 23 | 14 | 45.6 | 18.9 | 36.5 | — | |
| ForkNetResolution=(80, 80), Trained on=NYU, RGB-D based=false2020.03 | 36.2 | 93.8 | 29.2 | 18.9 | 17.7 | 61.6 | 52.9 | 23.3 | 19.5 | 45.4 | 20 | 37.1 | — | |
| ForkNet2019.09 | 36.2 | 93.8 | 29.2 | 18.9 | 17.7 | 61.6 | 52.9 | 23.3 | 19.5 | 45.4 | 20 | 37.1 | — | |
| ForkNetInput=D, Resolution=(80, 80)2021.12 | 36.2 | 93.8 | 29.2 | 18.9 | 17.7 | 61.6 | 52.9 | 23.3 | 19.5 | 45.4 | 20 | — | 0.371 | |
| ForkNetInputs=TSDF, Input depth condition=noisy depth2023.03 | 36.2 | 93.8 | 29.2 | 18.9 | 17.7 | 61.6 | 52.9 | 23.3 | 19.5 | 45.4 | 20 | 37.1 | — | |
| ForkNetablation=without scene consistency2019.09 | 36.8 | 91.7 | 28 | 18.3 | 8.3 | 58.8 | 49.5 | 13 | 16.7 | 42.6 | 17.6 | 34.7 | — | |
| SSCNetevaluation_protocol=observed surface evaluation2019.09 | 37.7 | 91.9 | 75.4 | 64 | 29 | 51.1 | 63.3 | 43.7 | 29.7 | 73.3 | 54.5 | 50.8 | — | |
| ForkNetevaluation_protocol=observed surface evaluation2019.09 | 41.5 | 90.8 | 69.6 | 54.8 | 27.7 | 53.1 | 66.3 | 44.4 | 27.1 | 74.7 | 57.5 | 55.2 | — | |
| 3D sketch-aware semantic scene completion frameworkResolution=(60, 60), Trained on=NYU, RGB-D based=true2020.03 | 43.1 | 93.6 | 40.5 | 24.3 | 30 | 57.1 | 49.3 | 29.2 | 14.3 | 42.5 | 28.6 | 41.1 | — | |
| 3D-SketchInput=RGB+D, Resolution=(60, 60)2021.12 | 43.1 | 93.6 | 40.5 | 24.3 | 30 | 57.1 | 49.3 | 29.2 | 14.3 | 42.5 | 28.6 | — | 0.411 | |
| 3D-SketchInputs=RGB+TSDF, Backbone=ResNet50, Input depth condition=noisy depth2023.03 | 43.1 | 93.6 | 40.5 | 24.3 | 30 | 57.1 | 49.3 | 29.2 | 14.3 | 42.5 | 28.6 | 41.1 | — | |
| IMENetInput=RGB+D, Resolution=(60, 60)2021.12 | 43.6 | 93.6 | 42.9 | 31.3 | 36.6 | 57.6 | 48.4 | 32.1 | 16 | 47.8 | 36.7 | — | 0.442 | |
| IMENetInputs=RGB+D, Input depth condition=noisy depth2023.03 | 43.6 | 93.6 | 42.9 | 31.3 | 36.6 | 57.6 | 48.4 | 32.1 | 16 | 47.8 | 36.7 | 44.2 | — | |
| FFNetInputs=RGB+D+TSDF, Input depth condition=noisy depth2023.03 | 44 | 93.7 | 41.5 | 29.3 | 36.2 | 59 | 51.1 | 28.9 | 26.5 | 45 | 32.6 | 44.4 | — | |
| CleanerSInputs=RGB+TSDF, Backbone=Segformer-B2, Input depth condition=noisy depth2023.03 | 46.3 | 93.9 | 43.2 | 33.7 | 38.5 | 62.2 | 54.8 | 33.7 | 39.2 | 45.7 | 33.8 | 47.7 | — | |
| CleanerSInputs=RGB+TSDF, Backbone=ResNet50, Input depth condition=noisy depth2023.03 | 48.2 | 94 | 41.5 | 29.9 | 34.4 | 60.7 | 49.7 | 30.9 | 33.9 | 42.7 | 31 | 45.2 | — | |
| PVA-NetInput=D, Resolution=(60, 60)2021.12 | 51.4 | 94 | 49.9 | 15.9 | 41.9 | 68.3 | 58.8 | 35.4 | 12.9 | 48.5 | 29.1 | — | 0.46 |