Research map / Image generation / Image restoration
Trash or Treasure? An Interactive Dual-Stream Strategy for Single Image Reflection Separation
Read the paperCheck 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 |
|---|---|---|---|
| add YTMT feature interaction between two streams (ablation) | dual-stream UNet · Real20 | PSNR: 20.57 -> 23.05 | Helped |
| add YTMT feature interaction between two streams (ablation) | dual-stream UNet · Postcard | PSNR: 20.82 -> 22.66 | Helped |
| add YTMT feature interaction between two streams (ablation) | dual-stream UNet · Wild | PSNR: 24.20 -> 25.24 | Helped |
| pass negative-ReLU (deactivated) features across streams instead of using ReLU only (ablation) | YTMT-UCS · Real20 | PSNR: 22.79 -> 23.05 | Helped |
| pass negative-ReLU features instead of ReLU only (ablation) | YTMT-UCS · Postcard | PSNR: 22.02 -> 22.66 | Helped |
| pass negative-ReLU features instead of ReLU only (ablation) | YTMT-UCS · Wild | SSIM: 0.859 -> 0.887 | Mixed |
| add second stage (S -> T) with concat blocks (ablation) | YTMT-UCS · Real20 | PSNR: 23.05 -> 23.09 | No clear effect |
| add second stage (S -> T) with concat blocks (ablation) | YTMT-UCS · Postcard | SSIM: 0.885 -> 0.884 | No clear effect |
| use feature addition instead of concatenation in single-stage YTMT (ablation) | YTMT-UCS · Wild | PSNR: 25.24 -> 24.49 | Hurt |
| use feature addition instead of concatenation in single-stage YTMT (ablation) | YTMT-UCS · Objects | PSNR: 24.46 -> 24.71 | Helped |
| YTMT-UCT vs IBCLN on test sets | IBCLN · Postcard | PSNR: 23.39 -> 22.91 | Hurt |
| YTMT-UCS vs IBCLN under IBCLN training setting | IBCLN · Nature | PSNR: 23.57 -> 23.85 | Helped |
Closest papers by meaning
- ReflexSplit: Single Image Reflection Separation via Layer Fusion-Separation (CVPR 2026 · similarity 0.90)
- Single Image Reflection Separation via Dual-Stream Interactive Transformers (NeurIPS 2024 · similarity 0.89)
- Revisiting Single Image Reflection Removal In the Wild (CVPR 2024 · similarity 0.86)
- Reversible Decoupling Network for Single Image Reflection Removal (CVPR 2025 · similarity 0.85)
- Language-guided Image Reflection Separation (CVPR 2024 · similarity 0.84)
- FIRM: Flexible Interactive Reflection ReMoval (AAAI 2025 · similarity 0.83)
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.