Skip to content

rafaca5tro/ml-cv-gfpgan

 
 

Repository files navigation

GFPGAN

Status Python PyTorch

Practical blind face restoration using Generative Facial Prior (CVPR 2021).

Restores degraded face images without paired training data. Includes Real-ESRGAN integration for background enhancement, pre-trained models, and Hugging Face Spaces demo.


Features

  • Blind face restoration (works on any degraded face)
  • Clean model (no CUDA extensions needed)
  • Background enhancement via Real-ESRGAN
  • Pre-trained model zoo
  • Command-line inference
  • Web demo (Gradio/Hugging Face)

Tech Stack

Layer Technology
Language Python >= 3.7
Framework PyTorch >= 1.7
Libraries basicsr, facexlib, OpenCV
Optional Real-ESRGAN

Quick Start

pip install -r requirements.txt
python inference_gfpgan.py -i inputs/ -o results/

License

Apache 2.0 (original), modified

About

GFPGAN aims at developing Practical Algorithms for Real-world Face Restoration.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages

  • Python 100.0%