Output with distortion

[ZYJGO]

Hallo there,

I would like to ask about what I have observed when training with Denoiser model. When I train with the DAPS dataset, I notice a significant distortion in the results, but almost no distortion when using the DNS dataset. Is this because the DAPS dataset has very small and uniformly noise? Or is it related to the model structure itself?

Hope to hear from you soon:)