Research map / Image generation / Editing and style transfer
ERL-Net: Entangled Representation Learning for Single Image De-Raining
Guoqing Wang, Changming Sun, Arcot Sowmya
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 skip-connections to the encoder-decoder de-raining model (ablation) | MEMD · Rain100H | PSNR: 24.0743 -> 29.3782 | Improved |
| add skip-connections to the encoder-decoder de-raining model (ablation) | MEMD · AGAN-Data (raindrop) | PSNR: 30.4712 -> 31.4305 | Improved |
| add the feature fusion layer after skip-connections (ablation) | MEMD · Rain100H | PSNR: 29.3782 -> 31.6310 | Improved |
| add the feature fusion layer after skip-connections (ablation) | MEMD · AGAN-Data (raindrop) | PSNR: 31.4305 -> 32.2139 | Improved |
| add the entangled residual encoder branch (MA D + MLGD) | MEMD · Rain100H | PSNR: 31.6310 -> 34.5724 | Improved |
| add the entangled residual encoder branch (MA D + MLGD) | MEMD · AGAN-Data (raindrop) | PSNR: 32.2139 -> 32.9647 | Improved |
| use multiple layer guided entanglement (MLGD) instead of middle-layer only (CLGD) (ablation) | ERL-Net MA D · Rain100H | PSNR: 33.0145 -> 34.5724 | Improved |
| use multiple layer guided entanglement (MLGD) instead of middle-layer only (CLGD) (ablation) | ERL-Net MA D · AGAN-Data (raindrop) | PSNR: 32.7124 -> 32.9647 | Improved |
| use stage-wise training (MA D) instead of direct one-stage training (ablation) | ERL-Net MLGD · Rain100H | PSNR: 31.6287 -> 34.5724 | Improved |
| use direct training of the residual branch without pre-training (ablation) | MEMD plus residual branch · Rain100H | PSNR: 31.6310 -> 31.6214 | No clear effect |
| add a third stage fine-tuning MDE with the pixel loss (MA C vs MA A) (ablation) | ERL-Net MLGD · Rain100H | PSNR: 31.6804 -> 33.2041 | Improved |
| fine-tune REN and MDE jointly in stage 3 instead of MDE only (MA D vs MA C) (ablation) | ERL-Net MLGD · Rain100H | PSNR: 33.2041 -> 34.5724 | Improved |
Closest papers by meaning
- Representation Entanglement for Generation: Training Diffusion Transformers Is Much Easier Than You Think (NeurIPS 2025 · similarity 0.89)
- Unsupervised Domain-Specific Deblurring via Disentangled Representations (CVPR 2019 · similarity 0.86)
- Multi-Decoding Deraining Network and Quasi-Sparsity Based Training (CVPR 2021 · similarity 0.85)
- Tripod: Three Complementary Inductive Biases for Disentangled Representation Learning (ICML 2024 · similarity 0.84)
- Rethinking Multi-Scale Representations in Deep Deraining Transformer (AAAI 2024 · similarity 0.84)
- Transferring disentangled representations: bridging the gap between synthetic and real images (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.
