Monocular Depth Estimation on Pano3D zero-shot GV2
86δ1 AccuracyUpper-bound
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Upper-boundbest solution per sample=true2026.05 | 86 | 0.132 | 0.36 | |
| UniK3D2026.05 | 82 | 0.152 | 0.383 | |
| DepthAgent2026.05 | 81.9 | 0.145 | 0.365 | |
| Rand-Sol2026.05 | 79.2 | 0.158 | 0.422 | |
| Rand-Model2026.05 | 78.7 | 0.16 | 0.429 | |
| UniK3DDataset Size=7.94M, Backbone=ViT-L [7], Zero-shot evaluation=true2026.03 | 78.5 | 0.17 | 0.4 | |
| UniDACDataset Size=1.45M, Backbone=ViT-L [7], Zero-shot evaluation=true2026.03 | 76.8 | 0.161 | 0.394 | |
| UniDAC2026.05 | 75.5 | 0.169 | 0.472 | |
| MLP2026.05 | 74.4 | 0.182 | 0.473 | |
| Max2026.05 | 73.5 | 0.235 | 0.5 | |
| DACUDataset Size=0.79M, Backbone=Swin-L [32], Zero-shot evaluation=true2026.03 | 68.4 | 0.203 | 0.507 | |
| Mean2026.05 | 62 | 0.199 | 0.529 | |
| Metric3Dv22026.05 | 42.2 | 0.3 | 0.847 | |
| UniDepth2026.05 | 41.5 | 0.314 | 0.898 | |
| Metric3Dv2Dataset Size=16.20M, Backbone=ViT-L [7], Zero-shot evaluation=true2026.03 | 40.4 | 0.307 | 0.855 | |
| Metric3D2026.05 | 40.3 | 0.29 | 0.721 | |
| DepthFusion2026.05 | 37.8 | 0.327 | 1 | |
| Min2026.05 | 26.6 | 0.354 | 1.091 | |
| UniDepthDataset Size=3.83M, Backbone=ViT-L [7], Zero-shot evaluation=true2026.03 | 24.7 | 0.789 | 0.127 |