V2 update

Filtered and updated the image dataset for more consistent output of clothing details and better support for flat style models.

Adding some images of undressing to the dataset allowed for the generation of some relatively intuitively unclothed states.

This lora is trained for one of the best stewardess, rosy cloud cheongsam. I've been on a flight with this airline and I was pretty impressed, but the characters in the training images don't seem to be from real flight attendants ahaha.

Recommended for use with 2.5D and photorealistic mods, the details of the ornamentation on the garments may be blurred, use upscale to improve the quality of the details. May generate the model's facial information in the training set, it is recommended to use adetailer to ensure that the facial output of the problem, maybe there will be a later version of the update.