Research map / 3D vision / Depth and stereo
Diffusion Models for Monocular Depth Estimation: Overcoming Challenging Conditions
Fabio Tosi, Pierluigi Zama Ramirez, Matteo Poggi
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 |
|---|---|---|---|
| fine-tune monocular networks on diffusion-generated night and rain images with depth pseudo-labels | DPT · nuScenes night | absRel: 0.354 -> 0.224 | Improved |
| fine-tune Depth Anything with diffusion-distilled data | Depth Anything · nuScenes night | absRel: 0.291 -> 0.219 | Improved |
| fine-tune ZoeDepth with diffusion-distilled data | ZoeDepth · nuScenes day-clear | RMSE: 8.517 -> 8.946 | Hurt |
| fine-tune md4all with diffusion-generated data instead of GAN day-to-night translation requiring target images | md4all-DD · nuScenes night | absRel: 0.242 -> 0.188 | Improved |
| fine-tune md4all-DD with diffusion data on RobotCar night | md4all-DD · RobotCar night | absRel: 0.391 -> 0.139 | Improved |
| use Depth Anything or DPT instead of md4all depth to condition the diffusion model (ablation) | md4all-DD ft. Ours · nuScenes day-rain | absRel: 0.147 -> 0.139 (Depth Anything); 0.143 (DPT) | Improved |
| fine-tune on generated adverse images from only daytime datasets (Mapillary, Cityscapes, KITTI, Apolloscape) | DPT · DrivingStereo day-rain | absRel: 0.188 -> 0.124 | Improved |
| fine-tune on daytime-only generated data | Depth Anything · DrivingStereo day-rain | absRel: 0.112 -> 0.110 | Improved |
| fine-tune on text-generated transparent and mirror objects instead of real ToM data and GT masks | DPT-Large · Booster ToM pixels | delta<1.25 (%): 96.17 vs 96.68 for Costanzino et al. | No clear effect |
| fine-tune Depth Anything on generated non-Lambertian objects | Depth Anything · Booster ToM pixels | delta<1.25 (%): 84.23 -> 98.91 | Improved |
Closest papers by meaning
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · similarity 0.91)
- Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation (CVPR 2024 · similarity 0.91)
- Pixel-Perfect Depth with Semantics-Prompted Diffusion Transformers (NeurIPS 2025 · similarity 0.91)
- More Than Generation: Unifying Generation and Depth Estimation via Text-to-Image Diffusion Models (NeurIPS 2025 · similarity 0.90)
- Depth Anything V2 (NeurIPS 2024 · similarity 0.89)
- Jasmine: Harnessing Diffusion Prior for Self-supervised Depth Estimation (NeurIPS 2025 · similarity 0.89)
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.
