Abstract

Face image inpainting is becoming more and more significant in the practical fields of criminal investigation, cosmetic plastic surgery, and security protection. But most of traditional existing work is based on diffusion and texture methods, failing to effectively use the embedding attribute knowledge and semantic information of face images. It is difficult to ensure the authenticity and preserve semantic consistency at the same time. To tackle this problem, a face image inpainting method based on attribute control is proposed. First, a modified generative adversarial network is proposed to restore face images and control facial features by adding additional face attribute information and a classification network. Then an attention module based on multi-scale feature fusion is proposed, through increasing the receptive field of the neural network to obtain more global features to improve face image inpainting quality without increasing the arithmetic complexity. Finally, combining the four evaluation metrics, MAE, PSNR, SSIM, and FID, the experiments on the CelebA dataset show that the proposed method can not only generate high-quality face image inpainting results steered by semantic information, but also generate corresponding facial features based on attribute control.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.