search

TencentARC / GFPGAN

GFPGAN aims at developing Practical Algorithms for Real-world Face Restoration.
Other
35.52k stars 5.89k forks source link

Problem about finetuning GFPGAN v1.4 #564

Open leonsylarfang opened 1 month ago

leonsylarfang commented 1 month ago

Hello Xintao,

We found that the direct inferencing based on GFPGAN v1.4 performs pretty well on our own datasets, whilst GFPGAN v1 inferencing is not high-quality.

However, when we tried to finetune this model, we found that only the related files of training GFPGAN v1 was released.

Could you please share the .pth files, including discriminator, with us? So l could finetune the GFPGAN v1.4 with our custom data.

Thank you very much!

Best Junheng

Afsalts commented 1 month ago

@leonsylarfang were u able to finetune

leonsylarfang commented 1 month ago

@leonsylarfang were u able to finetune I have not got a reply yet... Finetuning GFPGAN v1 is ok but the data shape of the else cannot meet the provided states, as the value of channel_multiplier has been changed to 2.

Afsalts commented 1 month ago

@leonsylarfang were u able to finetune I have not got a reply yet... Finetuning GFPGAN v1 is ok but the data shape of the else cannot meet the provided states, as the value of channel_multiplier has been changed to 2.

is there any way i could talk to you . i have some doubts

Tsai-chasel commented 3 weeks ago

能解决嘛？我看toRGB这层就有问题