CodeFormer

CodeFormer copied to clipboard

Mimahmed

mdalmi812

If I retrain the model on real scene images, can Codeformer perform inpainting on images of natural scenes? For example, the Imagenet data set

duxiangcheng

pre-trained VQGAN of 256*256 resolution ratio

3

It takes too long to train STAGE1(about 16days), so I want to add my own dataset for fine-tuning. Can you release pre-trained VQGAN model in STAGE1? Thanks a lot~

create-li

自己训练stage2，stage3模型，推理时，设置w不等于0和1时，嘴唇会出现色块，有点偏绿色，紫色等

3

请问，基于vqgan训练stage3后，推理时，w不设置0,1时，就会出现人嘴唇的出现色块，但是w=0,1时就没有问题。

create-li

lacolumna

Why do the restored face images seem more clear than those in the final results?

GIT-LZL629

Training log share

7

Thanks for your Great work! Recently, I'm trying to reproduce the result in the paper but the model performs badly after three-stage training. Would you pelease share your trianing log...

CongHuang96

resolve error: AttributeError: module 'PIL.Image' has no attribute 'ANTIALIAS' use Pillow version: 9.5.0 the later versions remove it 

weidehai

Issues while loading the stage2 model

18

First thank you for releasing all the codes and models! I have been attempting to load the stage2 model provided [here](https://github.com/sczhou/CodeFormer/releases/download/v0.1.0/codeformer_stage2.pth). However, It seems that the keys do not correspond...

DannyXYZa

Can you tell me what is the dataset format for training a model for CodeFormer? Or do you know any upscaler-like codeformer that does anime?

JamesKnight0001