How to inference without the Noisy HQ Image input (x^{HQ}_t)?

[o0t1ng0o]

Hi @JyChen9811 ,

Thank you for sharing your code.
I have read through you paper. The aligment module is proposed to align the noisy hq image and lq image features.
But the module seems only integrate them together instead of aligment between them.
Moreover, during inference, we do not have the HQ image. How can we get the nosiy hq image for aligment module?
So how to inference without the Noisy HQ Image input (x^{HQ}_t)?

Thank you very much.

Best,

[whitewz]

i have the same question