Research map / Image generation / Image restoration
Training Image Estimators without Image Ground Truth
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 |
|---|---|---|---|
| Train from paired measurements using swap- and self-measurement consistency instead of image ground truth | Image estimators for compressive reconstruction · BSD68 and Set11, 1%, 4%, and 10% compression | PSNR (dB): BSD68: 19.74 vs 19.67, 22.94 vs 22.78, 25.57 vs 25.40; Set11: 17.88 vs 17.84, 22.61 vs 22.20, 26.74 vs 26.33 | Mixed |
| Add the self-measurement loss (ablation) | Unsupervised compressive reconstruction · BSD68 and Set11, 1%, 4%, and 10% compression | PSNR (dB): BSD68: 19.59 -> 19.67, 22.73 -> 22.78, 25.32 -> 25.40; Set11: 17.80 -> 17.84, 22.10 -> 22.20, 26.16 -> 26.33 | Helped |
| Add proxy-image supervision (ablation) | Unsupervised non-blind deblurring · Helen face deblurring test images | PSNR: 25.47 -> 25.95 | Helped |
| Add proxy-image supervision (ablation) | Unsupervised non-blind deblurring · CelebA face deblurring test images | PSNR: 24.64 -> 25.09 | Helped |
| Add proxy-image supervision (ablation) | Unsupervised non-blind deblurring · Helen face deblurring test images | SSIM: 0.867 -> 0.878 | Helped |
| Add proxy-image supervision (ablation) | Unsupervised non-blind deblurring · CelebA face deblurring test images | SSIM: 0.873 -> 0.885 | Helped |
| Train blind with a learned blur-kernel estimator and proxy data | Unsupervised face deblurring · Helen and CelebA face deblurring test images | PSNR: Helen: 25.95 vs 25.93; CelebA: 25.09 vs 25.06 | No clear effect |
Closest papers by meaning
- Training Deep Learning Based Image Denoisers From Undersampled Measurements Without Ground Truth and Without Image Prior (CVPR 2019 · similarity 0.91)
- Compressive sensing with un-trained neural networks: Gradient descent finds a smooth approximation (ICML 2020 · similarity 0.90)
- Algorithmic Guarantees for Inverse Imaging with Untrained Network Priors (NeurIPS 2019 · similarity 0.88)
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 2024 · similarity 0.88)
- Cross-Scale Self-Supervised Blind Image Deblurring via Implicit Neural Representation (NeurIPS 2024 · similarity 0.87)
- A Restoration Network as an Implicit Prior (ICLR 2024 · similarity 0.86)
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.