Face2Makeup: Generating Facial Makeup Images via Diffusion Model
Zhirui Huang, Xiaoyu Chen, Hui Wei, Zhaoyin Su, Yatao Liu, Zesen Dong · 2024
In recent years, the advancement of artificial intelligence technologies and diffusion models has significantly propelled the field of facial image analysis, particularly in facial reconstruction applications. In response to these developments, we propose the Face2Makeup model, which incorporates an image compression module to reduce computational resource requirements during training. Additionally, the model utilizes the F+roop module to control facial similarity, ensuring that the generated makeup aligns more closely with the user's facial features. Furthermore, we designed a Face+ESRGAN high-definition module to enhance image resolution, providing users with more accurate facial detail references. Finally, we contribute the C-makeup-gallery, a high-resolution, fine-grained facial makeup dataset, to address the limitations of existing datasets. Face2Makeup demonstrates superior performance in the makeup domain compared to existing models, as evidenced by the PSNR metric.