Research map / Image generation / Image restoration
AutoDIR: Automatic All-in-One Image Restoration with Latent Diffusion
Yitong Jiang, Zhaoyang Zhang, Tianfan Xue, Jinwei Gu
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 semantic-agnostic regularization when fine-tuning CLIP for degradation detection (ablation) | SA-CLIP BIQA · Degradation detection (7 degradations + clean) | average accuracy: 83.31% -> 97.94% | Improved |
| use CLIP with language prompts instead of a ViT-only classifier (ablation) | BIQA stage · Degradation detection (7 degradations + clean) | average accuracy: 77.65% -> 83.31% | Improved |
| use original pre-trained CLIP without fine-tuning for degradation detection (ablation) | BIQA stage · Degradation detection (7 degradations + clean) | average accuracy: 61.96% vs 77.65% ViT-classifier | Hurt |
| add SA-BIQA guidance to the latent diffusion restoration (with SCM) (ablation) | AutoDIR · Deraindrop | PSNR: 28.84 -> 30.10 | Improved |
| add SCM to the latent diffusion restoration (with SA-BIQA) (ablation) | AutoDIR · Deraindrop | PSNR: 25.58 -> 30.10 | Improved |
| add SCM to the latent diffusion restoration (with SA-BIQA) (ablation) | AutoDIR · Dehaze | PSNR: 24.47 -> 29.34 | Improved |
| add SA-BIQA guidance (with SCM) (ablation) | AutoDIR · Dehaze | PSNR: 27.53 -> 29.34 | Improved |
| use small w (0.1) for the structural-correction module on super-resolution (ablation) | SC-LDM · Super-resolution | perceptual quality (MUSIQ) | Improved |
| use generative AutoDIR instead of discriminative PromptIR on super-resolution | AutoDIR · Super-resolution | PSNR: 29.62 -> 23.49 | Hurt |
| initialize desnowing fine-tuning from pre-trained all-in-one AIR weights (ablation) | AIR backbone · Unseen de-snowing (CSD, 90 pairs) | adaptation speed | Improved |
Later papers that compared with this method
6 later accepted papers report AutoDIR as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- RestoreVAR: Visual Autoregressive Generation for All-in-One Image Restoration (ICLR 2026 · better on 6 of 7 shared results)
- FoundIR-v2: Optimizing Pre-Training Data Mixtures for Image Restoration Foundation Model (CVPR 2026 · better on 5 of 7 shared results)
- PRISM: Controllable Diffusion for Compound Image Restoration with Scientific Fidelity (ICLR 2026 · better on 5 of 5 shared results)
- Restore, Assess, Repeat: A Unified Framework for Iterative Image Restoration (CVPR 2026 · better on 1 of 2 shared results)
- Residual Diffusion Bridge Model for Image Restoration (CVPR 2026 · better on 1 of 1 shared result)
- Learning Efficient and Effective Trajectories for Differential Equation-Based Image Restoration (TPAMI 2025 · better on 0 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PromptIR — PromptIR: Prompting for All-in-One Image Restoration (NeurIPS 2023) · better on 6 of 9 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Visual-Instructed Degradation Diffusion for All-in-One Image Restoration (CVPR 2025 · similarity 0.87)
- DreamClear: High-Capacity Real-World Image Restoration with Privacy-Safe Dataset Curation (NeurIPS 2024 · similarity 0.87)
- GenDeg: Diffusion-based Degradation Synthesis for Generalizable All-In-One Image Restoration (CVPR 2025 · similarity 0.87)
- RestoreAgent: Autonomous Image Restoration Agent via Multimodal Large Language Models (NeurIPS 2024 · similarity 0.86)
- Taming Generative Diffusion Prior for Universal Blind Image Restoration (NeurIPS 2024 · similarity 0.86)
- Adapting Text-to-Image Generation with Feature Difference Instruction for Generic Image Restoration (CVPR 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.
