/nsfw_data_scraper

Collection of scripts to aggregate image data for the purposes of training an NSFW Image Classifier

Primary LanguageShellMIT LicenseMIT

NSFW Data Scraper

Disclaimer: the data is noisy - do not use to train a production model

Description

This is a set of scripts that allows for an automatic collection of tens of thousands of images for the following (loosely defined) categories to be later used for training an image classifier:

  • porn - pornography images
  • hentai - hentai images, but also includes pornographic drawings
  • sexy - sexually explicit images, but not pornography. Think nude photos, playboy, bikini, etc.
  • neutral - safe for work neutral images of everyday things and people
  • drawings - safe for work drawings (including anime)

Here is what each script (located under scripts directory) does:

  • 1_get_urls.sh - iterates through text files under scripts/source_urls downloading URLs of images for each of the 5 categories above. The Ripme application performs all the heavy lifting. The source URLs are mostly links to various subreddits, but could be any website that Ripme supports. Note: I already ran this script for you, and its outputs are located in raw_data directory. No need to rerun unless you edit files under scripts/source_urls.
  • 2_download_from_urls.sh - downloads actual images for urls found in text files in raw_data directory.
  • 3_optional_download_drawings.sh - (optional) script that downloads SFW anime images from the Danbooru2018 database.
  • 4_optional_download_neutral.sh - (optional) script that downloads SFW neutral images from the Caltech256 dataset
  • 5_create_train.sh - creates data/train directory and copy all *.jpg and *.jpeg files into it from raw_data. Also removes corrupted images.
  • 6_create_test.sh - creates data/test directory and moves N=2000 random files for each class from data/train to data/test (change this number inside the script if you need a different train/test split). Alternatively, you can run it multiple times, each time it will move N images for each class from data/train to data/test.

Prerequisites

  • Java runtime environment:
    • debian and ubuntu:sudo apt-get install default-jre
  • Linux command line tools: wget, convert (imagemagick suite of tools), rsync, shuf

On Windows

  • option 1: download a linux distro from windows 10 store and run the scripts there

  • option 2

    • download and install git from here. Git also installs Bash on your pc
    • download and install wget from here and add it to PATH
    • run the scripts

On Mac

The only difference I encountered is that OS X does not have shuf command, but has gshuf instead that can be installed with brew install coreutils. After installation either create an alias from gshuf to shuf or rename shuf to gshuf in 6_create_test.sh.

How to run

Change working directory to scripts and execute each script in the sequence indicated by the number in the file name, e.g.:

$ bash 1_get_urls.sh # has already been run
$ find ../raw_data -name "urls_*.txt" -exec sh -c "echo Number of URLs in {}: ; cat {} | wc -l" \;
Number of URLs in ../raw_data/drawings/urls_drawings.txt:
   25732
Number of URLs in ../raw_data/hentai/urls_hentai.txt:
   45228
Number of URLs in ../raw_data/neutral/urls_neutral.txt:
   20960
Number of URLs in ../raw_data/sexy/urls_sexy.txt:
   19554
Number of URLs in ../raw_data/porn/urls_porn.txt:
  116521
$ bash 2_download_from_urls.sh
$ bash 3_optional_download_drawings.sh # optional
$ bash 4_optional_download_neutral.sh # optional
$ bash 5_create_train.sh
$ bash 6_create_test.sh
$ cd ../data
$ ls train
drawings hentai neutral porn sexy
$ ls test
drawings hentai neutral porn sexy

I was able to train a CNN classifier to 91% accuracy with the following confusion matrix: alt text

As expected, anime and hentai are confused with each other more frequently than with other classes.

Same with porn and sexy categories.

Note: anime category was later renamed to drawings

Download Trained Models

This repo provides you with the data. To train or download a trained model you can access both here: