Clothing designer tricks AI-powered face recognition into thinking you're an animal

adversarial-patchesfacial-recognitionprivacycomputer-visionyolo

Abstraction: Adversarial knitted clothing patterns evade real-time facial recognition

Key points:

Milan startup Cap_able released "Manifesto" collection using knitted patterns that cause facial recognition cameras to detect animals (dogs, zebras, giraffes) instead of humans
Technology uses adversarial patches — images that cause recognition systems to ignore real subjects and detect a specified decoy item
Specifically targets YOLO ("you only look once"), an open-source real-time object detection neural network considered the fastest in existence
Cameras in public places do not recognize wearers as people, preventing biometric data collection
Founded by 2019 FIT graduate Rachele Didero; prices range from $310 (crew) to $460 (hoodie)
Motivated by privacy rights concerns around misuse of biometric recognition cameras in public surveillance

Connections: Cap Able · Yolo · Adversarial Patches · Facial Recognition · Computer Vision