Predicted face problem in Stage1

[HualeiZhuu]

I trained 80000 steps in Stage1, and I found that the output faces were very strange, is it OK ? prediction: gt:

So I used the pre-trained model vqgan_code1024.pth to predict, the faces were also strange. prediction:

[hailsham]

Ｉmet the similar problem when input the following image using released vqgan_code1024.pth the output is

The brightness and color changes are too obvious？

[hailsham]

Ｉmet the similar problem when input the following image using released vqgan_code1024.pth the output is

The brightness and color changes are too obvious？

Ｉ found the problem. it caused by my wrong normalizion.

[HualeiZhuu]

Yes, the problem is the normalization. In function nondist_validation, I change sr_img = tensor2img([visuals['result']]) to sr_img = tensor2img([visuals['result']], rgb2bgr=True, min_max=(-1, 1)), then the output becomes normal.

[heorhiikalaichev]

How do you run inference for model from the first part?

[YunanwuNancy]

Ｉmet the similar problem when input the following image using released vqgan_code1024.pth the output is The brightness and color changes are too obvious？

Ｉ found the problem. it caused by my wrong normalizion.

Hi, could you please twll how to solve this bright issue? What did you change for the normalization? Thanks.