GPEN-Windows upscaler
Posted: Sun Sep 26, 2021 11:18 pm
Some times ago I created a package to allow users to use GPEN upscaler and enhancer without install any version of python, cuda, MVSC, Windows 10 Toolkit, ninja and other hard dependencies.
HOW TO INSTALL:
- Download package from Register to see link or Register to see link
- Extract the rar file (I compressed it with WinRAR 'cause basically 90% of users use it and RAR can compress more than zip (I used 512MB dictionary))
- Inside the root folder open 1) install_dependences.bat file
HOW TO USE:
- To enhance files you need to put them inside examples\imgs folder. You can put both *DFL and not DFL images.
- Open 2) start_gpen.bat to run the GPEN. A CMD window will open and remain without any message for a while; no problem, the program is just compiling external dependencies written in C++.
- Choose to use normal model (512x512) or lite one (256x256)
- After upscaling results will be inside examples\outs folder. They will have still DFL metadata inside (polys, XSeg mask, landmarks, others...)
OTHER FUNCTIONS:
GPEN has other models to do other things; one of this is the colorization from scale of grays to colors. To do this just put images in the right foldes.
Folders map:
Enhancing: 2) start_gpen.bat --> imgs --> outs
Colorization: 2) start_gpen_colorization.bat --> grays --> outs-colorization
Inpainting: 2) start_gpen_face_inpainting.bat --> impainting --> outs-inpainting
Segmentation2face: 2) start_gpen_segmentation2face.bat --> segs --> outs-seg2face
HOW TO INSTALL:
- Download package from Register to see link or Register to see link
- Extract the rar file (I compressed it with WinRAR 'cause basically 90% of users use it and RAR can compress more than zip (I used 512MB dictionary))
- Inside the root folder open 1) install_dependences.bat file
HOW TO USE:
- To enhance files you need to put them inside examples\imgs folder. You can put both *DFL and not DFL images.
- Open 2) start_gpen.bat to run the GPEN. A CMD window will open and remain without any message for a while; no problem, the program is just compiling external dependencies written in C++.
- Choose to use normal model (512x512) or lite one (256x256)
- After upscaling results will be inside examples\outs folder. They will have still DFL metadata inside (polys, XSeg mask, landmarks, others...)
OTHER FUNCTIONS:
GPEN has other models to do other things; one of this is the colorization from scale of grays to colors. To do this just put images in the right foldes.
Folders map:
Enhancing: 2) start_gpen.bat --> imgs --> outs
Colorization: 2) start_gpen_colorization.bat --> grays --> outs-colorization
Inpainting: 2) start_gpen_face_inpainting.bat --> impainting --> outs-inpainting
Segmentation2face: 2) start_gpen_segmentation2face.bat --> segs --> outs-seg2face