Research map / 3D vision / Depth and stereo
Towards Zero-Shot Scale-Aware Monocular Depth Estimation
Vitor Guizilini, Igor Vasiljevic, Dian Chen, Rareș Ambruș, Adrien Gaidon
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 input-level geometric embeddings (camera rays) to DPT (ablation) | DPT · KITTI zero-shot, no median scaling | AbsRel: 0.240 -> 0.144 | Improved |
| use 3D geometric embeddings instead of 2D positional embeddings (ablation) | ZeroDepth · KITTI zero-shot, no median scaling | AbsRel: 0.175 -> 0.100 | Improved |
| add intermediate variational latent representation on top of geometric-embedded DPT (ablation) | ZeroDepth · KITTI zero-shot, no median scaling | AbsRel: 0.144 -> 0.100 | Improved |
| use variational latent representation instead of deterministic Perceiver latent (ablation) | ZeroDepth · KITTI zero-shot, no median scaling | AbsRel: 0.122 -> 0.100 | Improved |
| add resolution jittering augmentation (ablation) | ZeroDepth · KITTI zero-shot, no median scaling | AbsRel: 0.142 -> 0.100 | Improved |
| add ray jittering augmentation (ablation) | ZeroDepth · KITTI zero-shot, no median scaling | AbsRel: 0.112 -> 0.100 | Improved |
| add embedding dropout (ablation) | ZeroDepth · KITTI zero-shot, no median scaling | AbsRel: 0.109 -> 0.100 | Improved |
| add surface normal regularization (ablation) | ZeroDepth · KITTI zero-shot, no median scaling | AbsRel: 0.109 -> 0.100 | Improved |
| train a single model on both indoor and outdoor data instead of separate models (ablation) | ZeroDepth · KITTI outdoor vs NYUv2 indoor | AbsRel: KITTI 0.100 -> 0.102; NYUv2 (metric) 0.104 -> 0.100 | Mixed |
| remove Waymo or LSD real-world training data (ablation) | ZeroDepth · KITTI zero-shot, no median scaling | AbsRel: 0.100 -> 0.133 (without LSD) | Hurt |
| keep only the 50% lowest-uncertainty pixels using 50 samples | ZeroDepth · KITTI zero-shot | RMSE: 4.044 -> 1.678 | Improved |
| reduce latent dimension D_l to 32 (ablation) | ZeroDepth · KITTI | RMSE (metric): 4.904 (median-scaled), 6.421 (metric) | Hurt |
Closest papers by meaning
- Metric from Human: Zero-shot Monocular Metric Depth Estimation via Test-time Adaptation (NeurIPS 2024 · similarity 0.91)
- UniDepth: Universal Monocular Metric Depth Estimation (CVPR 2024 · similarity 0.91)
- Depth Pro: Sharp Monocular Metric Depth in Less Than a Second (ICLR 2025 · similarity 0.90)
- RSA: Resolving Scale Ambiguities in Monocular Depth Estimators through Language Descriptions (NeurIPS 2024 · similarity 0.89)
- SharpDepth: Sharpening Metric Depth Predictions Using Diffusion Distillation (CVPR 2025 · similarity 0.89)
- Repurposing Marigold for Zero-Shot Metric Depth Estimation via Defocus Blur Cues (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.
