CV Dazzle is a form of camouflage from computer vision created in 2010 as my masters thesis at New York University’s Interactive Telecommunications Program. Unlike traditional camouflage, such as disruptive-pattern material, that hides the wearer from human observation, CV Dazzle is designed to break machine vision systems while still remaining perceptible to human observers. It is the first documented camouflage technique to successfully attack a computer vision algorithm.
Exposing.ai is a research project about the origins and endpoints of biometric image datasets created “in the wild”. The project investigates how photos have unwittingly become part of an information supply chain powering the global biometrics industry.
Research from the Exposing.ai project hsa been featured in the Financial Times, New York Times, Nature, a US Government Accountability Report, the 2020 AI Index, several academic research papers, and has helped pushed forward an urgent discussion about the ethics of dataset collection into public discourse.
The HyperFace (Version 1) prototype was developed for Hyphen-Labs NeuroSpeculative AfroFeminism project and debuted at the Sundance Film Festival in 2017. The project was collaboration with Hyphen Labs members Ashley Baccus-Clark, Carmen Aguilar y Wedge, Ece Tankal, Nitzan Bartov, and JB Rubinovitz.
Automatic, private, open-source face redaction web app: try it here
DFACE.app
DFACE uses the YOLOV5 neural network object detection framework to run face detection in a web browser so photos never leave a user’s device. It can process up to 1,000 faces per image at down to 10x10 pixels per face with varying effects (color fill, blur, or emoji), and supports batch-processing multiple images. It is designed for activists and social media users to quickly and privately redact faces in imagery before posting to social media.