Monocular Depth Estimation on ScanNet (100 sequences)
70Delta Accuracy 1Ensemble
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EnsembleBackbone=ResNet50-Monodepth22026.05 | 70 | 1.33 | 17 | 18 | 1,857 | 1,245 | |
| UfM-EnsembleBackbone=ResNet50-Monodepth22026.05 | 68 | 1.34 | 16 | 16 | 257 | 1,319 | |
| UfM-Ensemble-100KBackbone=ResNet50-Monodepth22026.05 | 68 | 1.51 | 17 | 17 | 112 | 1,249 | |
| UfM*-EnsembleBackbone=ResNet50-Monodepth22026.05 | 68 | 0.96 | 12 | 13 | 66 | 1,245 | |
| MC-DropoutBackbone=ResNet50-Monodepth22026.05 | 68 | 3.87 | 24 | 19 | 1,854 | 0.01 | |
| UfM-MCDBackbone=ResNet50-Monodepth22026.05 | 68 | 1.95 | 18 | 16 | 242 | 76 | |
| UfM-MCD-100KBackbone=ResNet50-Monodepth22026.05 | 68 | 2.33 | 20 | 17 | 111 | 4 | |
| UfM*-MCDBackbone=ResNet50-Monodepth22026.05 | 68 | 1.12 | 13 | 13 | 61 | 0.03 | |
| BatchEnsembleBackbone=ResNet50-Monodepth22026.05 | 68 | 2.15 | 22 | 17 | 1,580 | 2 | |
| AleatoricBackbone=ResNet50-Monodepth22026.05 | 68 | 4.1 | 24 | 19 | 10 | 0.01 | |
| UfM-AleatoricBackbone=ResNet50-Monodepth22026.05 | 68 | 1.99 | 18 | 16 | 234 | 76 | |
| UfM-Aleatoric-100KBackbone=ResNet50-Monodepth22026.05 | 68 | 2.38 | 20 | 17 | 109 | 4 | |
| UfM*-AleatoricBackbone=ResNet50-Monodepth22026.05 | 68 | 1.11 | 13 | 13 | 61 | 0.04 | |
| UfM-BatchEnsembleBackbone=ResNet50-Monodepth22026.05 | 67 | 1.48 | 18 | 15 | 302 | 80 | |
| UfM-BatchEnsemble-100KBackbone=ResNet50-Monodepth22026.05 | 67 | 1.74 | 20 | 16 | 162 | 6 | |
| UfM*-BatchEnsembleBackbone=ResNet50-Monodepth22026.05 | 67 | 1.01 | 11 | 12 | 123 | 2 | |
| EvidentialBackbone=ResNet50-Monodepth22026.05 | 67 | 3.51 | 22 | 16 | 30 | 0.03 | |
| UfM-EvidentialBackbone=ResNet50-Monodepth22026.05 | 67 | 1.85 | 19 | 14 | 245 | 75 | |
| UfM-Evidential-100KBackbone=ResNet50-Monodepth22026.05 | 67 | 2.17 | 20 | 15 | 119 | 4 | |
| UfM*-EvidentialBackbone=ResNet50-Monodepth22026.05 | 67 | 1.12 | 16 | 13 | 74 | 0.05 |