Research map / Label-efficient and robust learning / Knowledge distillation
QualiTeacher: Quality-Conditioned Pseudo-Labeling for Real-World Image Restoration
Fengyang Xiao, Jingjia Feng, Peng Hu, Yuhan Chen, Dingming Zhang, Lei Xu, Guanyi Qin, Lu Li, Chunming He, Sina Farsiu
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 quality-score conditioning to the student (ablation) | NAFNet mean-teacher · Desnowing RealSnow 10K-Test | HyperIQA: 0.452 -> 0.510 | Improved |
| add score-based preference optimization (ablation) | QualiTeacher · Desnowing | HyperIQA: 0.510 -> 0.535 | Improved |
| add cropped consistency loss (ablation) | QualiTeacher · Desnowing | HyperIQA: 0.535 -> 0.543 | Improved |
| add dual-drop mechanism (ablation) | QualiTeacher · Desnowing | HyperIQA: 0.543 -> 0.559 | Improved |
| QualiTeacher instead of Colabator on desnowing with NAFNet | NAFNet · Desnowing RealSnow 10K-Test | HyperIQA: 0.470 -> 0.559 | Improved |
| QualiTeacher instead of Colabator on desnowing with HistoFormer | HistoFormer · Desnowing RealSnow 10K-Test | LIQE: LIQE 2.429 vs 2.610 baseline; HyperIQA 0.480 -> 0.530 | Mixed |
| QualiTeacher instead of Colabator on deraining | HistoFormer · Deraining Real3000 | Topiq / LIQE: Topiq 0.520 -> 0.508; HyperIQA 0.545 -> 0.547 | Mixed |
| QualiTeacher on dehazing with CORUN | CORUN · Dehazing RTTS | FADE: 0.824 -> 0.590 | Improved |
| QualiTeacher on LLIE with CIDNet versus Colabator | CIDNet · LLIE DICM | MFRQA: 59.319 -> 59.341 | Improved |
| enhance low-light images with QualiTeacher before detection | CIDNet · ExDark detection | mean AP: 76.7 -> 79.1 | Improved |
| replace MUSIQ with TOPIQ in training IQA (ablation) | QualiTeacher · Desnowing | HyperIQA: 0.559 -> 0.538 | No clear effect |
Closest papers by meaning
- Semi-Supervised Blind Quality Assessment with Confidence-quantifiable Pseudo-label Learning for Authentic Images (ICML 2025 · similarity 0.85)
- Learning from History: Task-agnostic Model Contrastive Learning for Image Restoration (AAAI 2024 · similarity 0.85)
- IRPO: Boosting Image Restoration via Post-training GRPO (NeurIPS 2026 · similarity 0.85)
- Learning Where and What to Restore for Composite Image Restoration (NeurIPS 2026 · similarity 0.85)
- UARE: A Unified Vision-Language Model for Image Quality Assessment, Restoration, and Enhancement (CVPR 2026 · similarity 0.84)
- DreamClear: High-Capacity Real-World Image Restoration with Privacy-Safe Dataset Curation (NeurIPS 2024 · 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.
