Depth Estimation 벤치마크
Depth Estimation on ScanNetV2
absolute relative error 낮을수록 좋음
- 2020-03-19 — DELTAS: absolute relative error 0.0932
- 2025-02-26 — Distill Any Depth: absolute relative error 0.042
| Rank | Model | absolute relative error | Delta < 1.25 | Paper | Code | Year |
|---|---|---|---|---|---|---|
| 1 | Distill Any Depth | 0.042 | 0.980 | Distill Any Depth: Distillation Creates a Stronger Monocular Depth Estimator | Westlake-AGI-Lab/Distill-Any-Depth | 2025 |
| 2 | DELTAS | 0.0932 | 0.9287 | DELTAS: Depth Estimation by Learning Triangulation And densification of Sparse points | magicleap/DELTAS | 2020 |
| 3 | LeReS | 0.095 | – | Learning to Recover 3D Scene Shape from a Single Image | aim-uofa/AdelaiDepth | 2020 |
| 4 | Distill Any Depth | 0.042 | 0.980 | Distill Any Depth: Distillation Creates a Stronger Monocular Depth Estimator | Westlake-AGI-Lab/Distill-Any-Depth | 2025 |
| 5 | DELTAS | 0.0932 | 0.9287 | DELTAS: Depth Estimation by Learning Triangulation And densification of Sparse points | magicleap/DELTAS | 2020 |
| 6 | LeReS | 0.095 | – | Learning to Recover 3D Scene Shape from a Single Image | aim-uofa/AdelaiDepth | 2020 |