Academic capstone · Final-year B.E.
Next-Gen Industrial Safety System
A final-year capstone combining a custom IoT wearable with computer vision to monitor PPE compliance and worker vitals, with tiered emergency alerts.
The problem
On a factory floor, a missing helmet or a worker in distress is easy to miss. Safety monitoring that depends on someone noticing is monitoring that fails at the wrong moment.
What I built
- A computer-vision model detecting PPE compliance — helmet, vest and safety gear — from live camera feeds, flagging non-compliant workers.
- A wearable hardware layer: a microcontroller-based sensor node with firmware streaming telemetry to the backend.
- Worker-vitals monitoring through the wearable.
- A cloud-hosted dashboard for real-time monitoring.
- A multi-tier emergency alert workflow that escalates to supervisors when readings cross hazard thresholds.
My engineering contribution
- Built the system as my final-year B.E. capstone, across both the hardware and the Python software stack.
- Trained and deployed the PPE-detection model.
- Designed the wearable sensor node, its firmware and the alert-escalation workflow.
- Published the work as a research paper (details under outcomes).
Architecture and workflow
A microcontroller-based node worn by the worker reads vitals and streams telemetry.
Technical decisions
- Two independent sources — camera and wearable — so a blind spot in one does not blind the system.
- Alerts escalate in tiers rather than firing everything at everyone at once.
- Hazard thresholds, not a person watching a screen, trigger the escalation.
Outcomes
- Published as “Next-Gen Industrial Safety: AI-Driven Wearable Equipment Detection and Emergency Alert”, IJSART Vol. 11, Issue 11 (November 2025), ISSN 2395-1052.
- Accepted at ICIISM-2026.