Research map / Learning theory and algorithms / Prediction and decision losses
Learning to Autofocus
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 the learned model instead of the strongest contrast-based baseline | Focal-stack green-channel autofocus · Focal stack green-channel (I*) | MAE: 2.056 -> 1.600 | Helped |
| Use the learned model instead of the strongest stereo-matching baseline | Focal-stack dual-pixel autofocus · Focal stack dual-pixel (D*) | MAE: 2.731 -> 1.611 | Helped |
| Use the learned model instead of the best single-slice learned-depth baseline | Single-slice dual-pixel autofocus · Single-slice dual-pixel (D1) | MAE: 7.176 -> 2.235 | Helped |
| Add a second autofocus step (ablation) | Our Model · Multi-step D1 | MAE: 2.235 -> 1.931 | Helped |
| Add a second autofocus step (ablation) | Our Model · Multi-step I1 | MAE: 4.321 -> 2.855 | Helped |
| Use dual-pixel input instead of green-channel input (ablation) | Our single-slice model · Single-slice autofocus | MAE: 4.321 -> 2.235 | Helped |
| Train and evaluate on registered focal stacks (ablation) | Our Model · Focal stack green-channel (I*) | MAE: 1.600 -> 1.570 | Helped |
| Replace the original-camera test set with the other-phone test set (ablation) | Single-slice dual-pixel autofocus · Other-phone test set | MAE: 2.235 -> 2.912 | Mixed |
| Use low-light scenes instead of light scenes (ablation) | Our single-slice dual-pixel model · Single-slice dual-pixel (D1) | MAE: 2.654 -> 3.479 | Hurt |
Closest papers by meaning
- Stabilizing and Accelerating Autofocus with Expert Trajectory Regularized Deep Reinforcement Learning (CVPR 2025 · similarity 0.86)
- DualFocus: Depth from Focus with Spatio-Focal Dual Variational Constraints (NeurIPS 2025 · similarity 0.84)
- One-Step Event-Driven High-Speed Autofocus (CVPR 2025 · similarity 0.82)
- RobIA: Robust Instance-aware Continual Test-time Adaptation for Deep Stereo (NeurIPS 2025 · similarity 0.81)
- Multi-Focus Image Fusion via Explicit Defocus Blur Modelling (AAAI 2025 · similarity 0.81)
- Intensity-Robust Autofocus for Spike Camera (CVPR 2024 · similarity 0.81)
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.