Research map / Continual learning / Online and replay-based
Beyond the Static-World: Lifelong Learning for All-in-One Medical Image Restoration
Shihao Shan, Hongying Liu, Fanhua Shang, Liang Wan, Jingjing Deng
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 MIDAB to the static all-in-one baseline (ablation) | static baseline · Static all-in-one, avg PSNR | PSNR (dB): 34.3813 -> 34.6065 | Improved |
| add MIDAB to the naive lifelong baseline (ablation) | lifelong fine-tuning baseline · Lifelong, avg PSNR | PSNR (dB): 30.5142 -> 30.6874 | Improved |
| add experience replay on top of MIDAB (ablation) | lifelong + MIDAB · Lifelong, avg PSNR | PSNR (dB): 30.6874 -> 30.7694 | Improved |
| add AFC feature-level consolidation on top of MIDAB + ER (ablation) | lifelong + MIDAB + ER · Lifelong, avg PSNR | PSNR (dB): 30.7694 -> 31.2144 | Improved |
| add diversity loss to AFC (ablation) | ROME · Lifelong, avg PSNR | PSNR (dB): 31.2144 -> 32.3575 | Improved |
| replace pixel-level experience replay mixing with adaptive feature-level consolidation (ablation) | ROME · Lifelong, avg PSNR | PSNR (dB): 30.7694 -> 31.2144 | Improved |
| ROME versus AMIR in lifelong setting | ROME · Lifelong All-in-one, avg PSNR | PSNR (dB): 31.4139 -> 32.3575 | Improved |
| add MIDAB adversarial balancing to reduce gradient conflict (ablation) | ROME · Gradient conflict, CT vs PET | cosine similarity: -0.31 -> -0.265 | Improved |
| ROME versus baseline on forgetting of D1 | ROME · D1 test set across D1 to D5 | PSNR drop (dB): 2.69 dB lost -> 1.74 dB lost | Improved |
Closest papers by meaning
- UniH3: Unifying Hierarchical Homogeneity and Heterogeneity for All-in-One Medical Image Restoration (ECCV 2026 · similarity 0.89)
- LIRA: Lifelong Image Restoration from Unknown Blended Distortions (ECCV 2020 · similarity 0.87)
- Perturbating, Tuning, and Collaborating: Harnessing Vision Foundation Models for Single Domain Generalization on Medical Imaging (AAAI 2025 · similarity 0.85)
- Debiased All-in-one Image Restoration with Task Uncertainty Regularization (AAAI 2025 · similarity 0.85)
- Knowledge Transfer Scaling Laws for 3D Medical Imaging (NeurIPS 2026 · similarity 0.84)
- Advancing Multiple Instance Learning with Continual Learning for Whole Slide Imaging (CVPR 2025 · similarity 0.84)
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.
