The results seem less than ideal

[Lufffya]

trafficstars

My training images:

My regularization images, 100 sheets:

Results obtained from prompt "a man wears a sks hat in winter":

It seems that there is a gap between the results in the paper:https://dreambooth.github.io/ No hat exactly the same as that in the training images, did something go wrong?

[XinyangHan]

Yes!!! Indeed!!! Same for me!!! The reconstruction and editing ability has a gap with the exhibited result in paper!!!!!!!!!!

[HU-Qiqi]

I think that is because the hats in your training dataset are not the same hat. the samples of the training set in the paper is a same thing.