Research map / Image generation / Image restoration
Self2Self With Dropout: Learning Self-Supervised Denoising From Single Image
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 |
|---|---|---|---|
| Add dropout during training and test (ablation) | Self2Self · Set9, σ=25 | PSNR (dB): 23.88 -> 31.74 | Helped |
| Add dropout-based ensemble at test (ablation) | Self2Self · Set9, σ=25 | PSNR (dB): 29.92 -> 31.74 | Helped |
| Train on Bernoulli-sampled image instances (ablation) | Self2Self · Set9, σ=25 | PSNR (dB): 23.12 -> 31.74 | Helped |
| Use partial convolutions instead of standard convolutions in the encoder (ablation) | Self2Self · Set9, σ=25 | PSNR (dB): 31.26 -> 31.74 | Helped |
| Use Self2Self instead of (C)BM3D | Image denoising · Set9, σ=25 | PSNR (dB): 31.67 -> 31.74 | Helped |
| Use Self2Self instead of N2N | Image denoising · Set9, σ=100 | PSNR (dB): 26.45 -> 26.27 | Hurt |
| Use Self2Self instead of (C)DnCNN | Image denoising · BSD68 | PSNR (dB): 29.14 -> 28.70; 26.20 -> 25.92 | Hurt |
| Use Self2Self instead of (C)DnCNN | Real-world image denoising · PolyU | PSNR (dB): 37.55 -> 37.52 | Hurt |
| Use Self2Self instead of DIP | Image inpainting · Set11, 50%, 70%, and 90% dropped | PSNR (dB): 33.48 -> 35.14; 28.50 -> 31.06; 24.24 -> 25.91 | Helped |
Closest papers by meaning
- High-Quality Self-Supervised Deep Image Denoising (NeurIPS 2019 · similarity 0.89)
- Noise2Score: Tweedie’s Approach to Self-Supervised Image Denoising without Clean Images (NeurIPS 2021 · similarity 0.89)
- Zero-Shot Noise2Mean: Gap Minimization for Efficient Denoising from a Single Noisy Image (AAAI 2025 · similarity 0.88)
- Exploring Efficient Asymmetric Blind-Spots for Self-Supervised Denoising in Real-World Scenarios (CVPR 2024 · similarity 0.87)
- Positive2Negative: Breaking the Information-Lossy Barrier in Self-Supervised Single Image Denoising (CVPR 2025 · similarity 0.87)
- APR-RD: Complemental Two Steps for Self-Supervised Real Image Denoising (AAAI 2025 · 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.