Reproduce “Anisotropic Stroke Control for Multiple Artists Style Transfer” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xuanhong Chen, Xirui Yan, Naiyuan Liu, Ting Qiu, Bingbing Ni
The ASMA-GAN framework from Shanghai Jiao Tong University introduces an anisotropic stroke control mechanism that dynamically adjusts brushstroke size to preserve semantic details during artistic style transfer. This single model also effectively learns and applies distinct styles from multiple artists, yielding superior stylization with high semantic retention and accurate style reproduction.
Key insights
- Dynamically varying style-stroke size (anisotropic control) according to semantic importance is crucial for generating artistically convincing and semantically coherent stylizations, mimicking human painting techniques.
- A unified generative model can effectively learn abstract style knowledge and apply distinct styles from multiple artists by leveraging Conditional Instance Normalization for style injection and training on diverse portfolios.
- Utilizing multi-scale features within a Projection Discriminator significantly enhances its ability to discern fine textural characteristics and overall artistic qualities, guiding the generator to produce more realistic and stylistically accurate outputs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.