Crowd Density Estimation

Tech Stack

The Problem

• Event organisers had no real-time crowd density data, creating dangerous overcrowding risks at venues.
• Manual headcounts became unreliable and unsafe beyond a few hundred people in complex spaces.
• Traditional sensor-based monitoring required expensive fixed infrastructure impractical to deploy.
• No tool existed for retrospective crowd density analysis from existing recorded event footage.
• Venues faced legal liability exposure from inability to enforce and evidence safe occupancy limits.

Our Solution

• Built a deep learning density estimation system using CNNs for accurate real-time crowd monitoring.
• Implemented live heatmap visualisation overlaid on camera feeds showing density distribution in real time.
• Designed the system to work with existing CCTV infrastructure without any hardware modification.
• Created zone-based alerting triggering automated warnings when crowd density exceeds defined thresholds.
• Added a historical analysis dashboard for post-event density trend review and safety planning.

Impact

Provided real-time crowd monitoring with 94% counting accuracy, enabling proactive safety interventions.