Monocular Depth Estimation on NYU v2 (val)
0.251RMSEGRIN_FT_NI
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRIN_FT_NIIntrinsics=false2024.09 | 0.251 | 0.051 | |
| DepthAnythingIntrinsics=false2024.09 | 0.264 | 0.056 | |
| ZeroDepthIntrinsics=false2024.09 | 0.269 | 0.074 | |
| ZoeDepthIntrinsics=false2024.09 | 0.277 | 0.077 | |
| DMDIntrinsics=true2024.09 | 0.296 | 0.072 | |
| DINOv3Arch=7B/16, Probe Type=Linear Probing, Backbone Status=Frozen2026.02 | 0.309 | — | |
| Metric3DIntrinsics=true2024.09 | 0.31 | 0.083 | |
| Web-DINOArch=7B/14, Probe Type=Linear Probing, Backbone Status=Frozen2026.02 | 0.466 | — | |
| VersaViTArch=H/14, Probe Type=Linear Probing, Backbone Status=Frozen2026.02 | 0.473 | — | |
| SigLIP 2Arch=G/16, Probe Type=Linear Probing, Backbone Status=Frozen2026.02 | 0.494 | — | |
| MAEArch=H/14, Probe Type=Linear Probing, Backbone Status=Frozen2026.02 | 0.517 | — | |
| OpenCLIPArch=G/14, Probe Type=Linear Probing, Backbone Status=Frozen2026.02 | 0.541 | — | |
| Qwen2-VL-ViTArch=H/14, Probe Type=Linear Probing, Backbone Status=Frozen2026.02 | 0.541 | — | |
| PEcoreArch=G/14, Probe Type=Linear Probing, Backbone Status=Frozen2026.02 | 0.59 | — | |
| TuringViT-24LInput Res.=1024, Frozen encoder=true, Depth head=DPT-style2026.06 | 50.7 | — | |
| SigLIP2-LInput Res.=384, Frozen encoder=true, Depth head=DPT-style2026.06 | 51.6 | — | |
| MobileCLIP2-LInput Res.=224, Frozen encoder=true, Depth head=DPT-style2026.06 | 53.3 | — | |
| TuringViT-18LInput Res.=1024, Frozen encoder=true, Depth head=DPT-style2026.06 | 53.7 | — |