Research map / Label-efficient and robust learning / Domain adaptation and generalization
Focus on Defocus: Bridging the Synthetic to Real Domain Gap for Depth Estimation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use an estimated defocus map as an intermediate signal before depth prediction (ablation) | PoolAE focal-stack depth estimation · Synthetic Shape and Appearance | MSE: 0.031 -> 0.008; 0.034 -> 0.007 | Helped |
| Use an estimated defocus map as an intermediate signal before depth prediction (ablation) | PoolAE focal-stack depth estimation · Synthetic Wide DoF (Random) | MSE: 0.050 -> 0.078 | Hurt |
| Use global pooling to compare features across focal-stack inputs (ablation) | Direct focal-stack depth estimation · Synthetic Shape (Random) | MSE: 0.097 -> 0.036 | Helped |
| Use global pooling to compare features across focal-stack inputs (ablation) | Direct focal-stack depth estimation · Synthetic Appearance (Random) | MSE: 0.095 -> 0.039 | Helped |
| Use global pooling to compare features across focal-stack inputs (ablation) | Direct focal-stack depth estimation · Synthetic Shape (All) | MSE: 0.014 -> 0.031 | Hurt |
| Use global pooling to compare features across focal-stack inputs (ablation) | Direct focal-stack depth estimation · Synthetic Appearance (All) | MSE: 0.012 -> 0.034 | Hurt |
| Train on an out-of-focus focal-stack image instead of an all-in-focus image (ablation) | Single-image depth estimation · Synthetic Shape and Appearance | MSE: 0.049 -> 0.032; 0.050 -> 0.031 | Helped |
| Predict depth from a focal stack instead of one image (ablation) | Depth estimation · Mobile Depth | MSE: 0.184 -> 0.045 | Helped |
| Fine-tune the synthetic-trained model on NYU (ablation) | Proposed model · NYU Depth v2, less than 2m | RMSE: 0.272 -> 0.181 | Helped |
| Fine-tune the synthetic-trained model on NYU (ablation) | Proposed model · 7-Scenes, less than 2m | RMSE: 0.282 -> 0.189 | Helped |
| Fine-tune the synthetic-trained model on NYU (ablation) | Proposed model · NYU Depth v2, 45 degrees | RMSE: 0.073 -> 0.111 | Hurt |
Closest papers by meaning
- Blurry-Edges: Photon-Limited Depth Estimation from Defocused Boundaries (CVPR 2025 · similarity 0.88)
- Residual Diffusion Deblurring Model for Single Image Defocus Deblurring (AAAI 2025 · similarity 0.87)
- Multi-Focus Image Fusion via Explicit Defocus Blur Modelling (AAAI 2025 · similarity 0.86)
- BLURD: Benchmarking and Learning using a Unified Rendering and Diffusion Model (NeurIPS 2024 · similarity 0.86)
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · similarity 0.86)
- Zero-shot Depth Completion via Test-time Alignment with Affine-invariant Depth Prior (AAAI 2025 · similarity 0.86)
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.