Research map / Image generation / Image restoration
Lottery Prior: Randomized Neural Compression for Zero-Shot Inverse Problems
Haotian Wu, Di You, Pier Luigi Dragotti, Deniz Gunduz
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 |
|---|---|---|---|
| use Lottery Prior instead of ZS-NCD | ZS-NCD · Kodak24 AWGN sigma=15 | PSNR: 33.18 -> 34.29 | Improved |
| use Lottery Prior instead of ZS-NCD | ZS-NCD · Kodak24 Poisson alpha=25 | PSNR: 28.77 -> 29.55 | Improved |
| use Lottery Prior instead of ZS-NCD | ZS-NCD · Set13 Poisson alpha=15 | PSNR: 26.44 -> 26.35 | No clear effect |
| use Lottery Prior instead of DIP | DIP · Set5 4x super-resolution with noise | PSNR: 21.37 -> 25.87 | Improved |
| use Lottery Prior instead of DIP | DIP · Set5 random-mask inpainting rho=0.5 | PSNR: 31.27 -> 33.22 | Improved |
| use Lottery Prior instead of DIP | DIP · Set5 random-mask inpainting rho=0.8 | PSNR: 27.14 -> 28.51 | Improved |
| use Lottery Prior instead of DIP | DIP · AFHQ noisy JPEG removal | PSNR: 24.13 -> 25.54 | Improved |
| use Lottery Prior (no pretraining) instead of RAM with self-supervised finetuning | RAM (Pretrained) · Demosaicing | PSNR: 36.87 vs 32.22 | Hurt |
| use Lottery Prior instead of pretrained RAM without finetuning | RAM (Pretrained) · Demosaicing | PSNR: 31.38 -> 32.22 | Improved |
| use Lottery Prior instead of ZS-NCD for efficiency | ZS-NCD · Set11 denoising runtime | sec/1k steps: 96.91 -> 12.38 | Improved |
| use pixel-masking blind-spot training (ZS-N2S, S2S) on high-resolution images | ZS-N2S · Set13/Kodak24 AWGN | PSNR: 18.18 on Set13 sigma=15 | Hurt |
Closest papers by meaning
- A Restoration Network as an Implicit Prior (ICLR 2024 · similarity 0.88)
- Stochastic Deep Restoration Priors for Imaging Inverse Problems (ICML 2025 · similarity 0.88)
- FiRe: Fixed-points of Restoration Priors for Solving Inverse Problems (CVPR 2025 · similarity 0.87)
- InvFusion: Bridging Supervised and Zero-shot Diffusion for Inverse Problems (NeurIPS 2025 · similarity 0.87)
- How many measurements are enough? Bayesian recovery in inverse problems with general distributions (NeurIPS 2025 · similarity 0.87)
- Adapt and Diffuse: Sample-adaptive Reconstruction via Latent Diffusion Models (ICML 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.
