This repo contains over 6000 images of female anatomy for the purposes of captioning or training captioning models. There may also be applications in image generation training.

There is a list.txt included which lists the filenames for use with joycaption. I have also included a subdirectory containing txt captions generated by joycaption. These captions share the same filename as the parent image.

Open source datasets have a distinct lack of human anatomy and pornographic content, and this gap has seeped into a large number of downstream models. It is my hope that you will use this data to bridge that gap and improve nudity recognition for communities which may benefit from their inclusion, such as the blind or visually impaired.

EDIT: It has come to my attention that I used the wrong prompt format during generation of the included captions. It is very likely that there are captions generated which do not conform to the standards required for model training. I implore you to do your own research on the appropriate use of the batch-caption.py script provided by JoyCaption: https://github.com/fpgaminer/joycaption