gfpgan

GFP-GAN

Input

aligned

Shape : (1, 3, 512, 512)

for face detection

(Image from https://github.com/TencentARC/GFPGAN/blob/master/inputs/whole_imgs/10045.png)

Output

aligned

Shape : (1, 3, 512, 512)

for face detection

Requirements

This model requires additional module.

# Install facexlib - https://github.com/xinntao/facexlib
# We use face restoration helper in the facexlib package
pip3 install facexlib

# If you want to enhance the background (non-face) regions with Real-ESRGAN,
# you also need to install the realesrgan package
# Install basicsr - https://github.com/xinntao/BasicSR
pip3 install basicsr
pip3 install realesrgan

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 gfpgan.py

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 gfpgan.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 gfpgan.py --video VIDEO_PATH

If the input image is aligned, specify the --aligned option.

$ python3 gfpgan.py --aligned

If you want to use the facexlib module for face restoration in face detection mode, specify the --facexlib option.

$ python3 gfpgan.py --facexlib

If you upsampling the image, specify the --upscale option with scale value. In addition, if you with Real-ESRGAN, specify the --realesrgan option.

$ python3 gfpgan.py --upscale 2 --realesrgan

Reference

• GFPGAN
• FaceXLib

Framework

Pytorch

Model Format

ONNX opset=11

Netron

GFPGANv1.3.onnx.prototxt
retinaface_resnet50.onnx.prototxt