Drishti AI-Event Guardian: Real-time crowd safety system cuts response time 34%

Drishti AI-Event Guardian is a new intelligent crowd monitoring and emergency response framework from researchers Ritabrata Roy Choudhury, Arkajyoti Karmakar, and Rudra Pratap Mitra. The system integrates multimodal data from CCTV networks and UAV platforms, processed by deep learning models on Google Vertex AI. Core capabilities include real-time crowd density estimation using YOLOv8, spatiotemporal anomaly detection, and predictive crowd-flow modeling via gradient-boosted regression (five-minute forecasts with 8.3% MAPE). Four specialized modules handle missing person identification via facial recognition, medical emergency dispatch, a conversational AI chatbot for incident reporting, and intelligent guard reallocation that dynamically adjusts personnel based on density shifts.

Evaluated on two large-scale Indian events—Kumbh Mela and the RCB Victory Parade—the system achieved an anomaly detection F1-score of 0.91, facial recognition precision of 0.93, and median alert latency of just 111 milliseconds. The chatbot resolved 89% of incident filings without human operators, and the guard reallocation engine reduced responder deployment latency by 34% compared to manual methods. The results mark a shift from passive surveillance to proactive crowd intelligence, creating a scalable foundation for safety at gatherings ranging from local events to mega festivals.