What question did this study set out to answer?

This research aims to enhance style transfer by preserving structure while blending multiple artistic styles.

Artistic Image Style Transfer Based on Fusion Generative Adversarial Networks and VGG16

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial explores multi-style image synthesis in artistic images, suggesting improved visual blending capabilities.

This research aims to enhance style transfer by preserving structure while blending multiple artistic styles.
Proposed Fusion GAN with a dual-branch architecture.
Utilized MS-COCO for content and WikiArt for style images.
Incorporated a hybrid loss formulation with perceptual content and Gram-matrix-based style loss.
Achieved partial improvements in structural content retention but faced challenges with high FID scores (indicating less realism).
Quantitative evaluation showed low PSNR and SSIM values, suggesting further development is needed.
Qualitative results indicated effective stylistic blending, but identified key areas for model optimization.

Chen et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark