Research map / 3D vision / Depth and stereo
Marigold-DC: Zero-Shot Monocular Depth Completion with Guided Diffusion
Massimiliano Viola, Kevin Qu, Nando Metzger, Bingxin Ke, Alexander Becker, Konrad Schindler, Anton Obukhov
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add sparse-depth test-time guidance to a frozen pretrained Marigold depth estimator | Marigold · ScanNet zero-shot | MAE: 0.083 -> 0.017 (Marigold + LS vs Marigold-DC) | Improved |
| add 10-prediction median ensemble (ablation) | Marigold-DC · NYU-Depth V2 | MAE: 0.057 -> 0.048 | Improved |
| use min-max initialization instead of least-squares for scale and shift (ablation) | Marigold-DC · NYU-Depth V2 subset | MAE: 0.065 -> 0.055 | Improved |
| optimize depth latent directly instead of modifying the score function (ablation) | Marigold-DC · NYU-Depth V2 subset | MAE: 0.058 -> 0.055 | Improved |
| increase DDIM denoising steps (ablation) | Marigold-DC · NYU-Depth V2 subset | MAE | Improved |
| increase learning rate for depth latent beyond 0.05 or decrease below (ablation) | Marigold-DC · NYU-Depth V2 subset | MAE | Hurt |
| increase ensemble size beyond 10 (ablation) | Marigold-DC · NYU-Depth V2 subset | MAE | No clear effect |
| use Marigold-DC versus VPP4DC on denser VOID guidance and DDAD | Marigold-DC · DDAD front camera | MAE: 1.344 -> 2.364 (VPP4DC vs Marigold-DC); RMSE 6.781 -> 6.449 | Mixed |
| use Marigold-DC on very sparse guidance of 150 points versus propagation methods | Marigold-DC · VOID 150 | MAE: 0.492 -> 0.194 (NLSPN vs Marigold-DC) | Improved |
Closest papers by meaning
- Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation (CVPR 2024 · similarity 0.93)
- Zero-shot Depth Completion via Test-time Alignment with Affine-invariant Depth Prior (AAAI 2025 · similarity 0.92)
- Repurposing Marigold for Zero-Shot Metric Depth Estimation via Defocus Blur Cues (NeurIPS 2025 · similarity 0.90)
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · similarity 0.88)
- Depth Pro: Sharp Monocular Metric Depth in Less Than a Second (ICLR 2025 · similarity 0.88)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
