Research map / Image generation / Editing and style transfer
Iterative Feature Transformation for Fast and Versatile Universal Style Transfer
Tai-Yin Chiu, Danna Gurari
Read the paperCheck 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 |
|---|---|---|---|
| replace WCT/LST transformation with iterative feature transformation | autoencoder cascade · Photo-realistic style transfer (30 pairs) | SSIM: 0.4476 -> 0.5301 (vs WCT); 0.4429 -> 0.5301 (vs LST) | Improved |
| replace WCT/LST transformation with iterative feature transformation | autoencoder cascade · Photo-realistic style transfer (30 pairs) | FSIM: 0.7643 -> 0.8197 (vs WCT); 0.7658 -> 0.8197 (vs LST) | Improved |
| use iterative feature transformation instead of NST back-propagation optimization | style transfer · Speed, content image 1024x1024 | time (s): 108.94 -> 1.07 | Improved |
| use iterative feature transformation instead of WCT (niter=1) | style transfer · Speed, content image 256x256 | time (s): 1.15 -> 0.15 | Improved |
| use iterative feature transformation instead of WCT (niter=1) | style transfer · Speed, content image 1024x1024 | time (s): 1.63 -> 1.07 | Improved |
| increase niter from 1 to 2 (more style effect) (ablation) | iterative feature transformation · Speed, content image 1024x1024 | time (s): 1.07 -> 2.15 | Hurt |
| increase niter from 2 to 3 (ablation) | iterative feature transformation · artistic style transfer | convergence | No clear effect |
| use pre-trained AdaIN instead of iterative transformation | AdaIN · Speed, content image 256x256 | time (s): 0.15 -> 0.036 (AdaIN faster) | Hurt |
| use iterative transformation instead of NST | autoencoder cascade · Photo-realistic style transfer (30 pairs) | SSIM: 0.5644 -> 0.5301 | Hurt |
Closest papers by meaning
- Universal Style Transfer via Feature Transforms (NeurIPS 2017 · similarity 0.92)
- A Closed-Form Solution to Universal Style Transfer (ICCV 2019 · similarity 0.91)
- Learning Linear Transformations for Fast Image and Video Style Transfer (CVPR 2019 · similarity 0.90)
- Diversified Arbitrary Style Transfer via Deep Feature Perturbation (CVPR 2020 · similarity 0.89)
- HSI: A Holistic Style Injector for Arbitrary Style Transfer (CVPR 2025 · similarity 0.88)
- OmniStyle: Filtering High Quality Style Transfer Data at Scale (CVPR 2025 · similarity 0.87)
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.
