Depth-from-Defocus on NYUv2 (test)
99.2Delta 1 ThresholdDirect Optimization Approach (Ours)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Direct Optimization Approach (Ours)Supervision=Analytical2026.02 | 99.2 | 99.4 | 99.5 | 0.109 | 0.0084 | |
| DFF-DFVInput=focal stack, Paradigm=Supervised Learning2023.03 | 96.7 | 98 | 99 | 0.232 | 0.445 | |
| DFF-DFVSupervision=Supervised learning2026.02 | 96.7 | 98 | 99 | 0.232 | 0.445 | |
| DFF-FVInput=focal stack, Paradigm=Supervised Learning2023.03 | 95.6 | 97.9 | 98.8 | 0.285 | 0.47 | |
| DFF-FVSupervision=Supervised learning2026.02 | 95.6 | 97.9 | 98.8 | 0.285 | 0.47 | |
| OursInput=focal stack, Paradigm=Fully Self-Supervised Learning2023.03 | 95 | 97.9 | 98.7 | 0.325 | 0.17 | |
| Si et al.Supervision=Self-supervised2026.02 | 95 | 97.9 | 98.7 | 0.325 | 0.17 | |
| Focus-NetInput=focal stack, Paradigm=Self-sup w/ AIF2023.03 | 74.8 | 89.2 | 94.9 | 0.611 | 0.172 | |
| Focus-NetSupervision=Self-supervised w/ AIF2026.02 | 74.8 | 89.2 | 94.9 | 0.611 | 0.172 | |
| Defocus-NetInput=defocus, Paradigm=Self-sup w/ AIF2023.03 | 73.2 | 88.7 | 95.1 | 0.623 | 0.176 | |
| Defocus-NetSupervision=Self-supervised w/ AIF2026.02 | 73.2 | 88.7 | 95.1 | 0.623 | 0.176 | |
| Gur and WolfInput=in-focus, Paradigm=Self-sup w/ AIF2023.03 | 72 | 88.7 | 95.1 | 0.649 | 0.184 | |
| Gur & WolfSupervision=Self-supervised w/ AIF2026.02 | 72 | 88.7 | 95.1 | 0.649 | 0.184 | |
| Suwajanakorn, Hernandez, and SeitzInput=focal stack, Paradigm=Analytical Methods2023.03 | 68.8 | 80.2 | 91.7 | 0.95 | 0.25 | |
| Suwajanakorn et al.Supervision=Analytical2026.02 | 68.8 | 80.2 | 91.7 | 0.95 | 0.25 | |
| Moeller et al.Input=focal stack, Paradigm=Analytical Methods2023.03 | 67 | 77.8 | 91.2 | 0.985 | 0.263 | |
| Moeller et al.Supervision=Analytical2026.02 | 67 | 77.8 | 91.2 | 0.985 | 0.263 |