Research map / Image generation / Image restoration
Zero-Shot Noise2Noise: Efficient Image Denoising Without Any Data
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 |
|---|---|---|---|
| Use the symmetric loss instead of the one-directional loss (ablation) | ZS-N2N · Kodak24, Gaussian σ=25 | PSNR (dB): 28.65 -> 29.07 | Helped |
| Add the consistency loss (ablation) | ZS-N2N · Kodak24, Gaussian σ=25 | PSNR (dB): 28.01 -> 29.07 | Helped |
| Add residual loss (ablation) | ZS-N2N · Kodak24, Gaussian σ=25 | PSNR (dB): 17.93 -> 29.07 | Helped |
| Replace the lightweight network with a UNet (3.3M parameters) (ablation) | ZS-N2N · Kodak24, Gaussian σ=25 | PSNR (dB): 29.07 -> 21.01 | Hurt |
| Reduce the network to 4k parameters (ablation) | ZS-N2N · Kodak24, Gaussian σ=25 | PSNR (dB): 29.07 -> 28.66 | Hurt |
| Reduce the network to 2k parameters (ablation) | ZS-N2N · Kodak24, Gaussian σ=25 | PSNR (dB): 29.07 -> 28.28 | Hurt |
| Reduce the network to 1k parameters (ablation) | ZS-N2N · Kodak24, Gaussian σ=25 | PSNR (dB): 29.07 -> 28.07 | Hurt |
| Reduce the network to 500 parameters (ablation) | ZS-N2N · Kodak24, Gaussian σ=25 | PSNR (dB): 29.07 -> 27.71 | Hurt |
| Use ZS-N2N instead of BM3D | Kodak24, Gaussian σ=25 · Kodak24, Gaussian σ=25 | PSNR (dB): 29.02 -> 29.07 | Helped |
| Use ZS-N2N instead of S2S | Kodak24, Gaussian σ=50 · Kodak24, Gaussian σ=50 | PSNR (dB): 26.22 -> 24.81 | Hurt |
| Use ZS-N2N instead of S2S | Kodak24, Poisson λ=10 · Kodak24, Poisson λ=10 | PSNR (dB): 27.29 -> 24.92 | Hurt |
| Use ZS-N2N instead of the strongest listed baseline | McMaster18, Poisson λ=50 · McMaster18, Poisson λ=50 | PSNR (dB): 29.89 -> 30.36 | Helped |
Closest papers by meaning
- Zero-Shot Noise2Mean: Gap Minimization for Efficient Denoising from a Single Noisy Image (AAAI 2025 · similarity 0.91)
- Zero-shot Denoising via Neural Compression: Theoretical and algorithmic framework (NeurIPS 2025 · similarity 0.89)
- APR-RD: Complemental Two Steps for Self-Supervised Real Image Denoising (AAAI 2025 · similarity 0.88)
- Positive2Negative: Breaking the Information-Lossy Barrier in Self-Supervised Single Image Denoising (CVPR 2025 · similarity 0.88)
- Masked Pre-training Enables Universal Zero-shot Denoiser (NeurIPS 2024 · similarity 0.88)
- Zero-Shot Blind-spot Image Denoising via Implicit Neural Sampling (CVPR 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.