OnTrack

OnTrack control centre dashboard

Inspiration

Public transit security relies on reactive measures, often responding after incidents occur. We wanted to use AI-driven automation to detect threats in real time while also providing valuable transit data for better planning.

What it does

OnTrack analyzes live security camera feeds in ETS busses and LRTs to detect weapons, threats, and anomalies using AI. Additionally, it tracks occupancy counts to help with transit route forecasting and optimization.

How we built it

We used:

Machine Learning & Computer Vision – YOLOv3 & OpenCV for threat detection
Backend – Python & Flask for processing alerts
Frontend – React for the monitoring dashboard
Design & Prototyping – Figma for UI/UX
AI Enhancement – LLMs for context-aware decision-making

Challenges we ran into

Reducing false positives in weapon detection
Ensuring privacy compliance while using AI on surveillance feeds
Optimizing real-time processing on existing transit camera infrastructure

Accomplishments that we're proud of

Successfully integrated AI-driven threat detection with webcams (security cameras)
Implemented occupancy tracking to assist with transit route planning
Designed a simple and intuitive security dashboard for transit authorities

What we learned

The importance of bias mitigation in AI models to ensure fair detection
How to balance privacy concerns with security improvements
The potential of AI-driven analytics for transit optimization beyond security

What's next for OnTrack

Live pilot testing on select ETS routes
Enhancing AI accuracy with continuous training and real-world feedback
Expanding beyond security to offer crowd management insights for better transit planning
Exploring edge AI processing to reduce reliance on cloud computing