Research map / Image generation / Editing and style transfer
Deep Network Interpolation for Continuous Imagery Effect Transition
Xintao Wang, Ke Yu, Chao Dong, Xiaoou Tang, Chen Change Loy
Check 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 |
|---|---|---|---|
| Interpolate fine-tuned N20 and N60 models to handle unseen N30 noise | DnCNN DNI, w/o BN · LIVE1 N30* | PSNR (dB): 23.86 -> 29.84 | Improved |
| Interpolate fine-tuned N20 and N60 models to handle unseen N40 noise | DnCNN DNI, w/o BN · LIVE1 N40* | PSNR (dB): 21.90 -> 28.28 | Improved |
| Interpolate fine-tuned N20 and N60 models to handle unseen N50 noise | DnCNN DNI, w/o BN · LIVE1 N50* | PSNR (dB): 27.34 -> 27.67 | Improved |
| Interpolate fine-tuned N20 and N60 models to handle unseen N30 noise | DnCNN DNI, w/ BN · LIVE1 N30* | PSNR (dB): 24.42 -> 29.46 | Improved |
| Interpolate fine-tuned N20 and N60 models to handle unseen N40 noise | DnCNN DNI, w/ BN · LIVE1 N40* | PSNR (dB): 26.58 -> 28.08 | Improved |
| Interpolate fine-tuned N20 and N60 models to handle unseen N50 noise | DnCNN DNI, w/ BN · LIVE1 N50* | PSNR (dB): 27.44 -> 27.66 | Improved |
| Add batch normalization to the DNI denoiser (ablation) | DnCNN DNI · LIVE1 unseen noise levels N30*, N40*, N50* | PSNR (dB): 29.84 vs 29.46; 28.28 vs 28.08; 27.67 vs 27.66 | Mixed |
| Use DNI rather than pixel interpolation of output images | MSE-to-GAN super-resolution transition · Super-resolution imagery effects | artifact and texture separation | Improved |
| Fine-tune convolutional layers as well as instance-normalization parameters (ablation) | Style transfer from mosaic to fire · Style transfer imagery | fire-texture transfer | Improved |
| Fine-tune instance-normalization parameters only (ablation) | Style transfer from mosaic to fire · Style transfer imagery | color transformation | Improved |
| Replace parameter-space DNI with pixel-space interpolation | Depth-of-field transition · Photo translation | ghosting artifacts | Hurt |
Closest papers by meaning
- Deep Feature Interpolation for Image Content Changes (CVPR 2017 · similarity 0.86)
- FunkNN: Neural Interpolation for Functional Generation (ICLR 2023 · similarity 0.86)
- DeepStereo: Learning to Predict New Views From the World's Imagery (CVPR 2016 · similarity 0.86)
- Interpolating Convolutional Neural Networks Using Batch Normalization (ECCV 2018 · similarity 0.85)
- Homomorphic Interpolation Network for Unpaired Image-to-image Translation (TPAMI 2020 · similarity 0.85)
- Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit (ICLR 2024 · similarity 0.83)
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.
