Research map / Image generation / Image restoration
CycleISP: Real Image Restoration via Improved Data Synthesis
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 CycleISP-generated realistic training data and the proposed denoiser instead of the strongest listed baseline | DND RAW denoising · DND RAW denoising | PSNR: 48.89 -> 49.13 | Helped |
| Use CycleISP-generated realistic training data and the proposed denoiser instead of the strongest listed baseline | SIDD RAW denoising · SIDD RAW denoising | PSNR: 45.52 -> 52.41 | Helped |
| Use the proposed method instead of RIDNet | DND sRGB denoising · DND sRGB denoising | PSNR: 39.23 -> 39.56 | Helped |
| Use the proposed method instead of RIDNet | SIDD sRGB denoising · SIDD sRGB denoising | PSNR: 38.71 -> 39.52 | Helped |
| Add short skip connections (ablation) | CycleISP RAW2RGB · RAW2RGB clean reconstruction | PSNR: 23.22 -> 45.41 | Helped |
| Add the color correction branch (ablation) | CycleISP RAW2RGB · RAW2RGB clean reconstruction | PSNR: 33.58 -> 45.41 | Helped |
| Add channel attention (ablation) | CycleISP RAW2RGB · RAW2RGB clean reconstruction | PSNR: 45.08 -> 45.41 | Helped |
| Add spatial attention (ablation) | CycleISP RAW2RGB · RAW2RGB clean reconstruction | PSNR: 44.67 -> 45.41 | Helped |
| Add channel and spatial attention (ablation) | CycleISP RAW2RGB · RAW2RGB clean reconstruction | PSNR: 42.96 -> 45.41 | Helped |
| Use parallel CA and SA rather than sequential CA then SA (ablation) | CycleISP RAW2RGB DAB · RAW2RGB clean reconstruction | PSNR: 45.17 -> 45.41 | Helped |
| Use parallel CA and SA rather than sequential SA then CA (ablation) | CycleISP RAW2RGB DAB · RAW2RGB clean reconstruction | PSNR: 45.16 -> 45.41 | Helped |
Closest papers by meaning
- Toward Convolutional Blind Denoising of Real Photographs (CVPR 2019 · similarity 0.90)
- Unprocessing Images for Learned Raw Denoising (CVPR 2019 · similarity 0.90)
- Universal Denoising Networks : A Novel CNN Architecture for Image Denoising (CVPR 2018 · similarity 0.89)
- Benchmarking Denoising Algorithms With Real Photographs (CVPR 2017 · similarity 0.89)
- IDR: Self-Supervised Image Denoising via Iterative Data Refinement (CVPR 2022 · similarity 0.89)
- sRGB Real Noise Modeling via Noise-Aware Sampling with Normalizing Flows (ICLR 2024 · similarity 0.87)
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.