https://github.com/magicse/GFPGANv1.3-to-ncnn

The GFPGAN network consists of two networks. Actually GFPGAN and StyleGAN2
https://github.com/magicse/GFPGANv1.3-to-ncnn

Last synced: about 1 month ago
JSON representation

The GFPGAN network consists of two networks. Actually GFPGAN and StyleGAN2

• Host: GitHub
• URL: https://github.com/magicse/GFPGANv1.3-to-ncnn
• Owner: magicse
• Created: 2022-09-19T00:58:46.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2022-09-22T11:09:20.000Z (about 2 years ago)
• Last Synced: 2024-08-01T20:48:00.041Z (4 months ago)
• Language: Python
• Homepage:
• Size: 91.8 KB
• Stars: 36
• Watchers: 3
• Forks: 5
• Open Issues: 3
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome - magicse/GFPGANv1.3-to-ncnn - The GFPGAN network consists of two networks. Actually GFPGAN and StyleGAN2 (Python)

README

        
### The GFPGAN network consists of two networks. Actually GFPGAN and StyleGAN2

### Export GFPGAN network from GFPGANv1.3 to ncnn

1 Clone repository


2 Download GFPGANv1.3.pth in root dir of this project


3 Run - Export_PTH_to_PT_GFPGAN.py to convert pth to pt


4 Download pnnx from https://github.com/pnnx/pnnx and unpack to root dir of this project


5 Run pnnx 

```

pnnx.exe gfpganv1_clean_model.pt inputshape=[1,3,512,512]

```

6 Output will be gfpganv1_clean_model.ncnn.param and gfpganv1_clean_model.ncnn.bin


### Export GFPGAN network from GFPGANv1.3 to onnx

1 Clone repository


2 Download GFPGANv1.3.pth in root dir of this project


3 Create folder "pretrained" in root dir of project


4 Put any jpg file 512x512 in root dir with name 1.jpg


4 Run - Export_GFPGANv1_3_to_ONNX.py to convert pth to ONNX


### Export StyleGAN2 from GFPGANv1.3 to onnx

1 Clone repository


2 Create folder "pretrained" in root dir of project


3 Download GFPGANv1.3.pth in root dir of this project


4 Run Export_StyleGAN2GeneratorClean_to_ONNX.py