Research map / 3D vision / Depth and stereo
Self-supervised Monocular Depth Estimation: Let's Talk About The Weather
Kieran Saunders, George Vogiatzis, Luis J. Manso
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 |
|---|---|---|---|
| naively train Monodepth2 on weather-augmented data (ablation) | Monodepth2 · KITTI sunny test | Abs Rel: 0.115 -> 0.118 | Hurt |
| naive augmented training on bad-weather test (ablation) | Monodepth2 · KITTI bad weather | Abs Rel: 0.249 -> 0.140 | Improved |
| train in (unaugmented, augmented) pairs instead of augmented only (ablation) | Robust-Depth · KITTI bad weather | Abs Rel: 0.140 -> 0.155 | Hurt |
| add semi-augmented depth warping using unaugmented images (ablation) | Robust-Depth · KITTI bad weather | Abs Rel: 0.155 -> 0.135 | Improved |
| use unaugmented pose for warping augmented images (ablation) | Robust-Depth · KITTI bad weather | Abs Rel: 0.155 -> 0.138 | Improved |
| add bi-directional pseudo-supervision depth loss L_PS (ablation) | Robust-Depth · KITTI bad weather | Abs Rel: 0.155 -> 0.139 | Improved |
| add pseudo-supervised pose loss (ablation) | Robust-Depth · KITTI bad weather | Abs Rel: 0.155 -> 0.146 | Improved |
| combine all components with MonoViT backbone | Robust-Depth · NuScenes-Night | Abs Rel: 0.313 -> 0.276 (vs MonoViT) | Improved |
| train on weather augmentation then test on sunny/cloudy DrivingStereo | Robust-Depth (ResNet18) · DrivingStereo Sunny | Abs Rel: 0.177 -> 0.185 | Hurt |
| use positional augmentations only (ablation) | Robust-Depth · KITTI sunny | Abs Rel: 0.115 -> 0.111 | Improved |
| use individual positional augmentation (vertical, tile, erase, scale) alone (ablation) | Monodepth2 · KITTI sunny | Abs Rel: 0.120 vertical; 0.117 tile vs 0.115 baseline | No clear effect |
Closest papers by meaning
- Synthetic-to-Real Self-supervised Robust Depth Estimation via Learning with Motion and Structure Priors (CVPR 2025 · similarity 0.90)
- Weather-Conditioned Depth Anything (ECCV 2026 · similarity 0.89)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · similarity 0.87)
- On the Uncertainty of Self-Supervised Monocular Depth Estimation (CVPR 2020 · similarity 0.86)
- From-Ground-To-Objects: Coarse-to-Fine Self-supervised Monocular Depth Estimation of Dynamic Objects with Ground Contact Prior (CVPR 2024 · similarity 0.85)
- Improved Monocular Depth Prediction Using Distance Transform Over Pre-semantic Contours with Self-supervised Neural Networks (CVPR 2025 · similarity 0.85)
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.
