Research map / Image generation / Editing and style transfer
Learning Linear Transformations for Fast Image and Video Style Transfer
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 |
|---|---|---|---|
| Learn the style transformation from covariance features rather than computing the transform directly | Universal artistic style transfer · Artistic style transfer | user preference: 35.3% | Helped |
| Use a shallower encoder up to relu3_1 to preserve content feature affinity across frames | Linear transformation style transfer · Video style transfer | user preference for stability: 81.8% | Helped |
| Add the spatial propagation network after style transfer | Photo-realistic style transfer · Photo-realistic style transfer | structural distortion | Helped |
| Use whitened content as the SPN guidance instead of original colored content | Photo-realistic style transfer with SPN · Photo-realistic style transfer | preservation of transferred style | Helped |
| Use covariance matrices as transformation-network inputs instead of image or feature inputs (ablation) | Transformation module · Artistic style transfer | visual style fidelity | Helped |
| Use a shallower auto-encoder up to relu3_1 rather than relu4_1 (ablation) | Linear transformation style transfer · Artistic style transfer | content distortion | Helped |
| Combine style reconstruction losses from multiple VGG layers (ablation) | Linear transformation style transfer · Artistic style transfer | style detail coverage | Helped |
| Use image inputs instead of covariance-matrix inputs for the transformation CNN (ablation) | Transformation module · Artistic style transfer | visual style fidelity | Hurt |
| Use feature-map inputs instead of covariance-matrix inputs for the transformation CNN (ablation) | Transformation module · Artistic style transfer | visual style fidelity | Hurt |
| Train one shared CNN for content and style inputs instead of two independent CNNs (ablation) | Transformation module · Universal style transfer | training convergence | Hurt |
| Transfer content features from relu3_1 rather than deeper relu4_1 features (ablation) | Linear transformation style transfer · Artistic style transfer | style abstraction | Mixed |
Closest papers by meaning
- A Content Transformation Block for Image Style Transfer (CVPR 2019 · similarity 0.92)
- Universal Style Transfer via Feature Transforms (NeurIPS 2017 · similarity 0.92)
- Real-Time Neural Style Transfer for Videos (CVPR 2017 · similarity 0.91)
- Adjustable Real-time Style Transfer (ICLR 2020 · similarity 0.90)
- Arbitrary Video Style Transfer via Multi-Channel Correlation (AAAI 2021 · similarity 0.89)
- Semantix: An Energy-guided Sampler for Semantic Style Transfer (ICLR 2025 · similarity 0.89)
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.