Monocular Depth Estimation on ScanNet (val)
4.7AbsRelReChannel-9B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ReChannel-9BCategory=RGB-native field readout, Backbone=FLUX-Klein 9B2026.07 | 4.7 | 97.6 | |
| Edit2PercCategory=Generative dense prediction methods2026.07 | 4.9 | 97.3 | |
| ReChannel-4BCategory=RGB-native field readout, Backbone=FLUX-Klein 4B2026.07 | 5.8 | 96.6 | |
| Lotus-DCategory=Generative dense prediction methods2026.07 | 6 | 96.3 | |
| GeoWizardCategory=Generative dense prediction methods2026.07 | 6.1 | 95.3 | |
| GenPerceptCategory=Generative dense prediction methods2026.07 | 6.2 | 96.1 | |
| MarigoldReal training samples=None, Synthetic training samples=74K, Zero-shot=true, Ensemble=true2023.12 | 6.4 | 95.1 | |
| MarigoldCategory=Generative dense prediction methods2026.07 | 6.4 | 95.2 | |
| MarigoldReal training samples=None, Synthetic training samples=74K, Zero-shot=true, Ensemble=false2023.12 | 6.9 | 94.5 | |
| OmnidataReal training samples=11.9M, Synthetic training samples=310K, Zero-shot=true2023.12 | 7.5 | 93.6 | |
| HDNReal training samples=300K, Synthetic training samples=None, Zero-shot=true2023.12 | 8 | 93.9 | |
| DPTReal training samples=1.2M, Synthetic training samples=188K, Zero-shot=true2023.12 | 8.2 | 93.4 | |
| LeReSReal training samples=300K, Synthetic training samples=54K, Zero-shot=true2023.12 | 9.1 | 91.7 | |
| DiverseDepthReal training samples=320K, Synthetic training samples=None, Zero-shot=true2023.12 | 10.9 | 88.2 | |
| MiDaSReal training samples=2M, Synthetic training samples=None, Zero-shot=true2023.12 | 12.1 | 84.6 |