Research map / Label-efficient and robust learning / Knowledge distillation
Learning Multiple Adverse Weather Removal via Two-Stage Knowledge Learning and Multi-Contrastive Regularization: Toward a Unified Model
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 one unified model trained with two-stage knowledge learning, CKT, and MCR instead of the all-in-one baseline | Unified adverse-weather removal · SOTS (Haze) | PSNR: 30.49 -> 33.95 | Helped |
| Use one unified model trained with two-stage knowledge learning, CKT, and MCR instead of MPRNet | Unified adverse-weather removal · Rain 1400 (Rain) | PSNR: 31.36 -> 33.13 | Helped |
| Use one unified model trained with two-stage knowledge learning, CKT, and MCR instead of MPRNet | Unified adverse-weather removal · CSD (Snow) | PSNR: 29.68 -> 31.35 | Helped |
| Use multiple weather-specialized teachers rather than one clear-image teacher (ablation) | MSBDN student on mixed testing set · Mixed testing set | PSNR: 26.975 -> 30.507 | Helped |
| Use multiple weather-specialized teachers rather than one mixed-data teacher (ablation) | MSBDN student on mixed testing set · Mixed testing set | PSNR: 29.666 -> 30.507 | Helped |
| Add PFP without BFM to multi-teacher transfer (ablation) | MSBDN student on mixed testing set · Mixed testing set | PSNR: 30.507 -> 29.784 | Hurt |
| Add BFM to PFP-based multi-teacher transfer (ablation) | MSBDN student on mixed testing set · Mixed testing set | PSNR: 29.784 -> 31.668 | Helped |
| Add PFP and BFM to multi-teacher transfer (ablation) | MSBDN student on mixed testing set · Mixed testing set | PSNR: 30.507 -> 31.668 | Helped |
| Use contrastive regularization during training (ablation) | MSBDN student with CKT · Mixed testing set | PSNR: 31.668 -> 32.183 | Helped |
| Use SCR during KC and HCR during KE rather than SCR at both stages (ablation) | MSBDN student with CKT · Mixed testing set | PSNR: 32.183 -> 32.814 | Helped |
| Use SCR during KC and HCR during KE rather than HCR at both stages (ablation) | MSBDN student with CKT · Mixed testing set | PSNR: 32.264 -> 32.814 | Helped |
| Train in two stages, with KC followed by KE, rather than single-stage training (ablation) | MSBDN student on mixed testing set · Mixed testing set | PSNR: 32.061 -> 32.814 | Helped |
Closest papers by meaning
- Language-driven All-in-one Adverse Weather Removal (CVPR 2024 · similarity 0.89)
- Continuous Adverse Weather Removal via Degradation-Aware Distillation (CVPR 2025 · similarity 0.88)
- Semantic Segmentation in Multiple Adverse Weather Conditions with Domain Knowledge Retention (AAAI 2024 · similarity 0.87)
- Genuine Knowledge from Practice: Diffusion Test-Time Adaptation for Video Adverse Weather Removal (CVPR 2024 · similarity 0.85)
- Real-World Adverse Weather Image Restoration via Dual-Level Reinforcement Learning with High-Quality Cold Start (NeurIPS 2025 · similarity 0.84)
- Learning Real-World Image De-weathering with Imperfect Supervision (AAAI 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.