Inspiration

Crowded public spaces such as metro stations, festivals, shopping malls, and transport hubs are becoming increasingly vulnerable to sudden panic, aggressive behaviour, and stampede-like incidents. Even a few seconds of delay in recognising the problem can escalate chaos, resulting in injuries and public fear. Traditional CCTV systems only record incidents but cannot proactively alert officials. We were inspired to address this real-world safety challenge by developing VigilAI — an intelligent, privacy-preserving AI system that detects early triggers of danger and predicts risks before they escalate. Our vision is to enable faster intervention, reduce harm, and make public spaces safer and more resilient.

What it does

VigilAI is a probabilistic edge-AI model designed specifically for real-time behavioural risk prediction in large crowds. It analyses live CCTV streams locally on edge devices, ensuring continuous monitoring without requiring internet connectivity. The system detects early warning patterns such as aggressive gestures, sudden high-intensity motion, and panic propagation. Once a safety threshold is crossed, VigilAI automatically triggers alerts to security personnel, enabling them to respond promptly before a situation becomes dangerous.

Development & Learning

This project helped us understand the balance between speed, accuracy, and privacy in real-time public safety systems. We implemented probabilistic modelling, motion dynamics analysis, and aggressive-pattern recognition that can operate on limited computational resources. We also learned how to ensure robustness against noise, occlusions, and diverse environmental lighting — critical challenges in surveillance AI.

Impact

VigilAI strengthens public safety by: • Providing early alerts for faster response • Supporting safe mobility and protected crowd experiences • Reducing fatalities and panic outcomes in high-density environments It promotes sustainable deployment by utilising existing CCTV networks and processing data locally, thereby respecting public privacy.

Conclusion

VigilAI aims to transform CCTV systems from reactive evidence collectors into proactive life-saving safety sensors. With continued development, mentorship, and pilot opportunities, this project can evolve into a scalable solution protecting millions of people every day. Together, we can prevent chaos — before it begins.

Built With

  • cctv
  • computer-vision
  • edge-computing
  • opencv
  • probabilistic-ml-models
  • python
  • tensorflow
Share this project:

Updates

Submission history