CFA-GAN: Cross fusion attention and frequency loss for image style transfer

Xiangtian Zheng, Xiaolin Yang, Qun Zhao, Hao Zhang, Xiufeng He, Jinsong Zhang, Xuan Zhang

Abstract: Highlights•We propose Cross-Fusion Attention (CFA) for image style transfer, addressing slow inference speed and loss of local details in GAN-based methods.•Our approach incorporates a Frequency Loss to bridge frequency domains and enhance training effectiveness.•A joint loss function aids GAN training, improving the image synthesis effect.