Research map / Image generation / Editing and style transfer
Controlling Perceptual Factors in Neural Style Transfer
Leon A. Gatys, Alexander S. Ecker, Matthias Bethge, Aaron Hertzmann, Eli Shechtman
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 |
|---|---|---|---|
| Add spatial guidance with eroded region masks and a global guidance channel | Neural Style Transfer · Spatially guided image stylisation | region-specific stylisation quality | Improved |
| Use spatial guidance to match corresponding regions in content and style images | Neural Style Transfer · Sky and ground stylisation | cross-region texture artefacts | Improved |
| Replace guided Gram Matrices with the guided-sums representation (ablation) | Spatially guided Neural Style Transfer · Spatially guided image stylisation | texture quality | Hurt |
| Preserve content-image colours by transferring style in luminance only | Neural Style Transfer · Colour-preserving stylisation | content-colour preservation | Improved |
| Use luminance-only transfer instead of colour matching (ablation) | Neural Style Transfer · Styles with prominent brushstrokes | brushstroke colour naturalness | Hurt |
| Match style-image colours to content-image colours before style transfer | Neural Style Transfer · Colour-preserving stylisation | colour preservation | Improved |
| Replace full-image style transfer with luminance-only transfer (ablation) | Neural Style Transfer · Colour-preserving stylisation | luminance-colour dependency preservation | Hurt |
| Construct a style image by transferring fine-scale style onto coarse-scale style | Neural Style Transfer · Style mixing across spatial scales | independent scale control | Improved |
| Use low-resolution stylisation followed by upsampling and high-resolution refinement | Neural Style Transfer · High-resolution stylisation | high-resolution optimisation iterations: 2.5 times fewer iterations | Improved |
| Use a luminance-trained feed-forward network instead of combining luminance output from a regular network with content colour (ablation) | Fast Neural Style Transfer · Colour-preserving stylisation | integration of stylisation with content structures | Improved |
| Train spatially guided feed-forward stylisation with fixed guidance masks, then use arbitrary masks at test time | Fast Neural Style Transfer · Spatially guided stylisation | guidance-mask generalisation | Improved |
Closest papers by meaning
- MetaStyle: Three-Way Trade-off among Speed, Flexibility, and Quality in Neural Style Transfer (AAAI 2019 · similarity 0.89)
- Style-Aware Normalized Loss for Improving Arbitrary Style Transfer (CVPR 2021 · similarity 0.88)
- Rethinking Style Transfer: From Pixels to Parameterized Brushstrokes (CVPR 2021 · similarity 0.88)
- Neural Preset for Color Style Transfer (CVPR 2023 · similarity 0.88)
- Universal Style Transfer via Feature Transforms (NeurIPS 2017 · similarity 0.88)
- Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer (CVPR 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.
