Research map / Image generation / Image restoration
Deep Decoder: Concise Image Representations from Untrained Non-convolutional Networks
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 |
|---|---|---|---|
| Fit the untrained underparameterized Deep Decoder to the noisy image | Deep Decoder · Denoising — image examples in Table | PSNR | Mixed |
| Use Deep Decoder instead of bicubic interpolation | Superresolution · Superresolution — image examples in Table | PSNR: 26.3 -> 26.4 | Helped |
| Use Deep Decoder instead of DIP | Superresolution · Superresolution — image examples in Table | PSNR | Mixed |
| Use Deep Decoder instead of DIP | Inpainting · Inpainting — image examples in Table | PSNR: 35.6 vs 32.3 | Mixed |
| Use 1x1 pixelwise channel combinations instead of 3x3 convolutions at approximately equal parameter count (ablation) | Deep Decoder · Image reconstruction | representation error: about 1dB | Helped |
| Apply upsampling before ReLU rather than after it (ablation) | Deep Decoder · Image representation | representation error | Helped |
| Use linear upsampling rather than convex nonlinear upsampling (ablation) | Deep Decoder · One-dimensional signal representation | representation quality | Helped |
| Use channel count k=32 rather than k=64 or k=128 at about 14dB noisy-image PSNR (ablation) | Deep Decoder denoising · Denoising | PSNR: about 14dB | Helped |
| Use channel count k=64 or k=128 rather than k=32 at about 20dB noisy-image PSNR (ablation) | Deep Decoder denoising · Denoising | PSNR: around 20dB | Helped |
| Add early stopping as regularization (ablation) | Deep Decoder denoising · Denoising | MSE: 5000 iterations; 50,000 iterations | Helped |
Closest papers by meaning
- Universal Denoising Networks : A Novel CNN Architecture for Image Denoising (CVPR 2018 · similarity 0.90)
- Robust And Interpretable Blind Image Denoising Via Bias-Free Convolutional Neural Networks (ICLR 2020 · similarity 0.89)
- Network Deconvolution (ICLR 2020 · similarity 0.89)
- Denoising and Regularization via Exploiting the Structural Bias of Convolutional Generators (ICLR 2020 · similarity 0.89)
- Non-Local Color Image Denoising With Convolutional Neural Networks (CVPR 2017 · similarity 0.88)
- Generalized Deep Image to Image Regression (CVPR 2017 · similarity 0.88)
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.