Research map / Image generation / Image restoration
Generalized Deep Image to Image Regression
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 |
|---|---|---|---|
| Use 3-branch RBDN instead of DCGRF | Denoising · BSD300 and PASCAL VOC2012 test images, σ=25–55 | PSNR: 28.77 vs 28.67 (σ=25); 25.22 vs 24.90 (σ=55) | Helped |
| Use 3-branch RBDN instead of DCGRF | Denoising · BSD300 and PASCAL VOC2012 test images, σ=60 | PSNR: 23.25 vs 24.45 | Hurt |
| Add recursive branching to the base network (ablation) | RBDN relighting · CMU-MultiPIE validation set | log reconstruction error | Helped |
| Use learnable upsampling instead of fixed bilinear upsampling (ablation) | RBDN relighting · CMU-MultiPIE validation set | log reconstruction error | Helped |
| Increase branch count from 0 to 1 (ablation) | RBDN relighting · CMU-MultiPIE validation set | relighting quality | Helped |
| Increase branch count from 1 to 3 (ablation) | RBDN relighting · CMU-MultiPIE validation set | relighting quality | Helped |
| Increase branch count from 3 to 4 (ablation) | RBDN relighting · CMU-MultiPIE validation set | relighting quality | Hurt |
| Increase base-network depth up to 9 convolution layers (ablation) | RBDN relighting · CMU-MultiPIE | relighting performance | Helped |
| Increase base-network depth beyond 9 convolution layers (ablation) | RBDN relighting · CMU-MultiPIE | relighting performance | No clear effect |
| Increase filter size, channel count, or transform filter size (K, c, T) (ablation) | RBDN relighting · CMU-MultiPIE | relighting performance | Helped |
| Use MSE loss instead of softmax cross-entropy with class rebalancing (ablation) | RBDN colorization · MS-COCO test set | color saturation | Hurt |
| Use softmax cross-entropy with class rebalancing instead of MSE (ablation) | RBDN colorization · MS-COCO test set | color saturation | Helped |
Closest papers by meaning
- Deep Decoder: Concise Image Representations from Untrained Non-convolutional Networks (ICLR 2019 · similarity 0.88)
- Deeply-Recursive Convolutional Network for Image Super-Resolution (CVPR 2016 · similarity 0.87)
- Network Deconvolution (ICLR 2020 · similarity 0.87)
- Image Super-Resolution via Deep Recursive Residual Network (CVPR 2017 · similarity 0.87)
- Universal Denoising Networks : A Novel CNN Architecture for Image Denoising (CVPR 2018 · similarity 0.86)
- Refining Architectures of Deep Convolutional Neural Networks (CVPR 2016 · similarity 0.85)
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.