categorizes and gives images a NSFW evaluation
input: url or image directory
output: categorized images and a NSFW evaluation and exports info to an SQLite database
SFWpy is very machine and environment-specific at the moment; it relies on paths for one particular machine and it's terribly dependent on environmental packages (i.e. it won't work on your machine without massive configuration)
For directories: the script reads in the images
For URLs: the script scrapes the site using ImageScraper and reads in the images.
Images are first processed by Inception in Tensorflow, a neural network for categorizing images with good accuracy. Then the images are processed through a Docker container supporting Caffe, the engine required for using Yahoo's open_nsfw evaluation.
Lastly, imagename/location as well as the categorization and NSFW evaluation are sent to an SQLite database
--To do:--
Flag images in db that score .7 or higher as NSFW for review
Add ability to change name of project
Work on interoperability -- may have to start over :/