Physics-informed Machine Learning for Robust Pedestrian Detection in Embedded Low-power Systems

Explore a novel approach to urban surveillance and pedestrian detection in this 44-minute talk by Cristian Axenie at the Alan Turing Institute. Delve into the challenges of population growth in urban areas and the need for efficient security solutions. Discover how physics-informed machine learning algorithms, combined with neuromorphic sensors and computing systems, offer a promising alternative to traditional camera-based surveillance. Learn about the key advantages of these systems, including high energy efficiency, fast data acquisition, local processing, improved data protection, and rational resource use. Gain insights into a cutting-edge real-time machine learning algorithm infused with physical motion models, capable of accurately detecting and tracking pedestrians and bikers in urban scenarios, both day and night. Understand the potential benefits of implementing such systems at a town-scale for improved urban safety and resource management.