This is one of multiple tools I'm working on as a Black Mirror-style proof-of-concept tool to demonstrate the ongoing harm of racial profiling, and how it may be perpetuated by face recognition and machine learning.
This repository shows a series of scripts I used to:
- scrape the names of the SS high command from the Wikipedia page
scrape-names.pyand output them as a .csv - use them to create folders
create-named-folders.py - finds portraits of their faces with serpapi/google search image api
scraper.py - extracts metadata for each name and saves it all as a .json file
wiki.py - some elementary face extraction in
haar_cascade.py
AryScraper is the scraper that created the male and female face "Aryan" data sets using photos of Holocaust perpetrators and the SS high command. Note: This whole exercise is arcane and not rooted in science but the interest was to use AI to see if we could perform some hyperblic "categorization" or averaging on samples of self-proclaimed "Aryans".
Such gestures in public take after "tactical media" or "tactical technology".
Although images here were compiled from publicly available sources, the dataset nor model is available here to prevent harmful misuse.
.
├── Makefile
├── .github // github actions
├── README.md
├── requirements.txt
├── __pycache__
├── test
├── automate_scraping.py
├── create_folders.py
├── haar_cascade.py
├── names.py
├── scrape_names.py
├── scraper.py
├── ss-ranks.csv
├── util.py
├── wiki.py
├── men // images of male faces
└── women // images of female faces