An intelligent crowd monitoring system for large public events will be developed. The system will be based on a multi-fidelity
approach: cellular networks will be used as large-scale coarse estimates of the crowd density, while Wifi access points deployed
through the event area will provide small-scale estimations. By combining the two estimates, the system will be able (i) to provide
crowd density and directional fluxes maps in real-time; (ii) to anticipate abnormal crowd movements, especially for safety issues,
by learning crowd dynamics.