Search results
Results from the WOW.Com Content Network
GFPGAN aims at developing a Practical Algorithm for Real-world Face Restoration. It leverages rich and diverse priors encapsulated in a pretrained face GAN ( e.g. , StyleGAN2) for blind face restoration.
GFPGAN. like. 840. Running on A10G. Discover amazing ML apps made by the community.
Run GFPGAN created by tencentarc, the #1 AI model for Practical Face Restoration. Restore old photos or AI generated faces with GFPGAN.
Thanks to the powerful generative facial prior and delicate designs, our GFP-GAN could jointly restore facial details and enhance colors with just a single forward pass, while GAN inversion methods require expensive image-specific optimization at inference.
This Generative Facial Prior (GFP) is incorporated into the face restoration process via novel channel-split spatial feature transform layers, which allow our method to achieve a good balance of realness and fidelity.
Thanks to the powerful generative facial prior and delicate designs, our GFP-GAN could jointly restore facial details and enhance colors with just a single forward pass, while GAN inversion methods require expensive image-specific optimization at inference.
GFPGAN is a blind face restoration algorithm towards real-world face images. It leverages the generative face prior in a pre-trained GAN ( e.g. , StyleGAN2) to restore realistic faces while precerving fidelity.