Research map / 3D vision / Depth and stereo
One Scene, Two Depths: Probing Geometric Ambiguity in Monocular Foundation Models
Xiaohao Xu, Feng Xue, Xiang Li, Haowei Li, Shusheng Yang, Tianyi Zhang, Matthew Johnson-Roberson, Xiaonan Huang
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 |
|---|---|---|---|
| query a frozen model with Laplacian-transformed input in addition to RGB | DAv2-L · MD-3k Reverse | ML-SRA: 0 -> 52.2 | Improved |
| query a frozen model with RGB/LVP pair instead of a duplicated single map | DAv2-L · MD-3k Overall | ML-SRA: 56.4 -> 75.5 | Improved |
| feed LVP input instead of RGB | DAv2-L · MD-3k Reverse | SRA(1): 65.3 -> 15.9 | Improved |
| use Laplacian prompt instead of Gaussian low-pass prompt (ablation) | DAv2-L · MD-3k Reverse | ML-SRA: 4.5 -> 52.2 | Improved |
| use Laplacian prompt instead of Gaussian low-pass prompt (ablation) | DAv2-L · MD-3k Same | ML-SRA: 98.2 -> 93.6 | Hurt |
| use Laplacian prompt instead of Canny binary edges (ablation) | DAv2-L · MD-3k Reverse | ML-SRA: 49.4 -> 52.2 | Improved |
| use Laplacian prompt instead of Canny binary edges (ablation) | DAv2-S · MD-3k Reverse | ML-SRA: 24.2 -> 36.9 | Improved |
| use Laplacian prompt instead of Sobel (ablation) | DAv2-L · MD-3k Reverse | ML-SRA: 47.2 -> 52.2 | Improved |
| use Laplacian prompt instead of Sobel (ablation) | DAv2-S · MD-3k Reverse | ML-SRA: 36.9 vs 40.3 | No clear effect |
| use Laplacian prompt instead of wavelet (ablation) | DAv2-L · MD-3k Reverse | ML-SRA: 50.4 -> 52.2 | Improved |
| increase DAv2 model size from S to L under RGB/LVP pair (ablation) | DAv2 · MD-3k Reverse | ML-SRA: 36.9 -> 52.2 | Improved |
| apply LVP instead of RGB input on non-ambiguous data | DAv2-L · ETH3D | AbsRel: 0.136 -> 0.177 | Hurt |
Closest papers by meaning
- Geometric Distillation from Rectified Stereo: Leveraging Epipolar Cues for Monocular Depth (ECCV 2026 · similarity 0.91)
- DepthFocus: Controllable Depth Estimation for See-Through Scenes (CVPR 2026 · similarity 0.90)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.89)
- Feat2GS: Probing Visual Foundation Models with Gaussian Splatting (CVPR 2025 · similarity 0.89)
- RefineAny3D: Depth Refinement as Semantic Alignment for Monocular 3D Detection (NeurIPS 2026 · 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.
