PPE Detection & Safety Compliance Monitor
A YOLOv8-based inspection assistant that flags PPE compliance in real time and powers instant field reports.
Completed · Course project
Developer
- PPE classes
- 9
- Live input support
- Webcam
- Compliance logic
- Safe / unsafe
Problem Statement
Construction supervisors needed a lightweight way to verify whether workers were wearing mandatory protective equipment.
Manual inspection was inconsistent, and existing tools struggled with multiple classes simultaneously.
Solution
- Trained YOLOv8 models to detect nine PPE categories (helmet, vest, gloves, boots, and their absence states).
- Added a compliance engine that labels frames as SAFE or UNSAFE based on detected combinations.
- Packaged the pipeline inside a Streamlit app capable of processing webcam feeds, uploaded videos, or still images.
Features
- Real-time inference with configurable confidence thresholds and per-class alerting.
- Exportable compliance logs for auditing and incident follow-up.
- Hardware-friendly deployment footprint suitable for field laptops.
Result
The project became the capstone deliverable for DPL302m and a foundation for future industrial safety experiments at SpeedyLabX.