Familiar Face Identification on MCUs - A Privacy-Preserving Solution for Personalizing Devices

Watch a technical conference talk exploring the development of privacy-focused familiar face identification technology for microcontrollers. Learn how Plumerai achieved secure, energy-efficient user recognition directly on devices through advanced model compression and training techniques involving three neural networks for object detection, face representation, and face matching. Discover practical implementation challenges, view live demonstrations, and understand how this innovation enables personalized experiences across everyday devices like TVs, doorbells, coffee machines and cars - all while maintaining user privacy by avoiding cloud computing. Gain insights into making AI more accessible for tiny devices through end-to-end deep learning approaches that balance hardware constraints with accuracy requirements.