Research map / Architectures and efficiency / CNN design
Rethinking and Improving the Robustness of Image 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 |
|---|---|---|---|
| Apply SWAG activation smoothing | pre-trained ResNet stylization · User preference study | preference (%): 17.0 -> 83.0 | Helped |
| Apply SWAG activation smoothing | pre-trained Inception-v3 stylization · User preference study | preference (%): 34.7 -> 65.3 | Helped |
| Apply SWAG activation smoothing | pre-trained Wide ResNet stylization · User preference study | preference (%): 32.3 -> 67.7 | Helped |
| Apply SWAG activation smoothing | pre-trained VGG stylization · User preference study | preference (%): 48.3 -> 51.7 | Helped |
| Apply SWAG activation smoothing | randomly initialized ResNet stylization · User preference study | preference (%): 15.7 -> 84.3 | Helped |
| Apply SWAG activation smoothing | randomly initialized Inception-v3 stylization · User preference study | preference (%): 10.7 -> 89.3 | Helped |
| Apply SWAG activation smoothing | randomly initialized Wide ResNet stylization · User preference study | preference (%): 5.4 -> 94.6 | Helped |
| Apply SWAG activation smoothing | randomly initialized VGG stylization · User preference study | preference (%): 44.3 -> 55.7 | Helped |
| Apply SWAG smoothing to the reconstruction loss (ablation) | pre-trained ResNet image reconstruction · 10 randomly selected content images | PSNR: 27.9(± 0.1) -> 28.3(± 0.1) | Helped |
| Remove residual connections (ablation) | ResNet style-transfer feature extractor · Image style transfer | stylization quality | Helped |
| Add residual connections back to pseudo-VGG (ablation) | pseudo-VGG style-transfer feature extractor · Image style transfer | stylization quality | Hurt |
| Add a 3×3 VGG convolution kernel, basic blocks, and stage maxpooling (ablation) | NoResNet style-transfer feature extractor · Image style transfer | stylization quality | Helped |
Closest papers by meaning
- Arbitrary Style Transfer With Deep Feature Reshuffle (CVPR 2018 · similarity 0.89)
- CAP-VSTNet: Content Affinity Preserved Versatile Style Transfer (CVPR 2023 · similarity 0.89)
- In the Light of Feature Distributions: Moment Matching for Neural Style Transfer (CVPR 2021 · similarity 0.88)
- Neural Style Transfer via Meta Networks (CVPR 2018 · similarity 0.88)
- StyleBank: An Explicit Representation for Neural Image Style Transfer (CVPR 2017 · similarity 0.88)
- Puff-Net: Efficient Style Transfer with Pure Content and Style Feature Fusion Network (CVPR 2024 · 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.