Gray setting ? #196

New Issue

claunia · 2026-01-29T21:45:34Z

claunia commented

2026-01-29 21:45:34 +00:00

Originally created by @gravitydeep on GitHub (Jun 10, 2022).

Linux 20.04
rtx 3060

Hello is that someone to explain me clearly how to configure the file .yml to colorize photos that are in black and white (gray) and then what is the command to launch because I am a beginner and I do not understand at all except the command that is on the rest to improve the quality, can you give me examples please thank you

Originally created by @gravitydeep on GitHub (Jun 10, 2022). Linux 20.04 rtx 3060 Hello is that someone to explain me clearly how to configure the file .yml to colorize photos that are in black and white (gray) and then what is the command to launch because I am a beginner and I do not understand at all except the command that is on the rest to improve the quality, can you give me examples please thank you

claunia commented

2026-01-29 21:45:35 +00:00

@AlionSSS commented on GitHub (Sep 2, 2023):

Maybe, you should use GFPGANv1.pth. Please see https://github.com/TencentARC/GFPGAN#european_castle-model-zoo.
Or you can use RestoreFormer. Here is an example command: python inference_gfpgan.py -i inputs/my_faces -o results/my_faces -v RestoreFormer -s 2 --suffix fixedRF
Picture example of RestoreFormer :

For full picture, you should use GFGAN with some colorization libs, such as DeOldify, colorization, colornet
Picture example of GFGAN V1.4 + colorization:

@AlionSSS commented on GitHub (Sep 2, 2023): - Maybe, you should use [GFPGANv1.pth](https://github.com/TencentARC/GFPGAN/releases/download/v0.1.0/GFPGANv1.pth). Please see https://github.com/TencentARC/GFPGAN#european_castle-model-zoo. - Or you can use [RestoreFormer](https://github.com/wzhouxiff/RestoreFormer). Here is an example command: `python inference_gfpgan.py -i inputs/my_faces -o results/my_faces -v RestoreFormer -s 2 --suffix fixedRF` - Picture example of **RestoreFormer** : ![Paris_Hilton_crop_00](https://github.com/TencentARC/GFPGAN/assets/20226368/d8b27c1d-40d1-42b0-bf1e-5610fd8a9c74) --- - For **full picture**, you should use GFGAN with some colorization libs, such as [DeOldify](https://github.com/jantic/DeOldify), [colorization](https://github.com/richzhang/colorization), [colornet](https://github.com/pavelgonchar/colornet) - Picture example of **GFGAN V1.4 + colorization**: ![Paris_siggraph17](https://github.com/TencentARC/GFPGAN/assets/20226368/bf94e6be-2aff-4d05-9499-54e1c5fb596c)

claunia commented

2026-01-29 21:45:35 +00:00

@luchaoqi commented on GitHub (Apr 13, 2024):

For full picture, you should use GFGAN with some colorization libs, such as DeOldify, colorization, colornet

Is this a separate step after the GFPGAN processing @AlionSSS ? Wondering if there's any easy way to use v1 for black and white images and v1.4 for regular images.

@luchaoqi commented on GitHub (Apr 13, 2024): > For full picture, you should use GFGAN with some colorization libs, such as [DeOldify](https://github.com/jantic/DeOldify), [colorization](https://github.com/richzhang/colorization), [colornet](https://github.com/pavelgonchar/colornet) Is this a separate step after the GFPGAN processing @AlionSSS ? Wondering if there's any easy way to use v1 for black and white images and v1.4 for regular images.

claunia commented

2026-01-29 21:45:35 +00:00

@AlionSSS commented on GitHub (Apr 13, 2024):

For full picture, you should use GFGAN with some colorization libs, such as DeOldify, colorization, colornet

Is this a separate step after the GFPGAN processing @AlionSSS ? Wondering if there's any easy way to use v1 for black and white images and v1.4 for regular images.

Yes

@AlionSSS commented on GitHub (Apr 13, 2024): > > For full picture, you should use GFGAN with some colorization libs, such as [DeOldify](https://github.com/jantic/DeOldify), [colorization](https://github.com/richzhang/colorization), [colornet](https://github.com/pavelgonchar/colornet) > > Is this a separate step after the GFPGAN processing @AlionSSS ? Wondering if there's any easy way to use v1 for black and white images and v1.4 for regular images. Yes

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: TencentARC/GFPGAN#196