alex000kim/nsfw_data_scraper ? reverse-engineered prompt

Build me a simple tool that can collect and organize images for an NSFW image classifier.

I want it to gather a big dataset into five folders, porn, hentai, sexy, neutral, and drawings, then clean it up and split it into train and test sets. It should be easy to run from Docker, since I just want one command that starts the whole process and leaves me with a ready dataset in data/train and data/test.

Please include scripts that can pull image links from source lists, download the actual images, optionally grab extra neutral and anime style images from public datasets, and remove broken files along the way. It should also copy all good jpg and jpeg images into the training folder, then move a random sample into the test folder.

If it helps, add a notebook or simple guide for training a basic classifier after the data is collected. Look up current docs online if you need to.