Used 333 images of women in various states. The results is a LoRA that enhances hand and feet anatomy, which is already good in many Illustrious models. The LoRA also seems to help produce higher quality poses that are more varied. About half of the images in the dataset are just straight up blatant foot fetish material and the rest are generated hand references, photo headshots, bodyshots and poses of various kinds. There are intentionally a mix of generated illustration style imagery and photography, and hopefully varied enough to avoid overfitting realism. I hope this was successful. Dataset images were generated with Midjourney and gathered from Pinterest.
Please have a good day <3