Depth Completion on ScanNet (AbsRel Metrics)
1AbsRel (SIFT)CAPALoRA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CAPALoRABackbone=VGGT, Adaptation strategy=LoRA2026.02 | 1 | 0.9 | 1.1 | |
| CAPAVPTBackbone=VGGT, Adaptation strategy=VPT2026.02 | 1.1 | 1 | 1 | |
| OMNI-DC-v1.1Estimator type=Depth completion2026.02 | 2 | 1.5 | 5.4 | |
| VGGTEstimator type=Multi-view reconstruction2026.02 | 2.2 | 2 | 2 | |
| UniDepthV2Estimator type=Single-image depth estimation2026.02 | 3.2 | 2.7 | 2.8 | |
| PriorDA-v1.1Estimator type=Depth completion2026.02 | 3.4 | 2.5 | 10.3 | |
| MoGe-2Estimator type=Single-image depth estimation2026.02 | 4.1 | 3.3 | 3.4 | |
| Marigold-DCEstimator type=Depth completion2026.02 | 4.3 | 3.7 | 2.2 | |
| TestPromptDCEstimator type=Depth completion2026.02 | 4.4 | 3.8 | 2.8 | |
| DepthAnythingV2Estimator type=Single-image depth estimation2026.02 | 4.8 | 3.8 | 4.7 | |
| VideoDAEstimator type=Video-based depth estimation2026.02 | 5 | 4.1 | 4.2 | |
| PromptDAEstimator type=Depth completion2026.02 | 16.6 | 13.2 | 12.2 |