Project Overview
SiteSafeTrack is an AI-powered safety compliance monitoring system designed for construction environments. Utilizing a custom-trained YOLO12 model, it detects and tracks essential Personal Protective Equipment (PPE) such as helmets and safety vests in real time. This system helps ensure workers adhere to safety protocols by automating the surveillance process through computer vision.
Tech Stack
-
Programming Language: Python
-
Deep Learning Model: YOLO12 (
Ultralyticslibrary) -
Libraries: OpenCV, NumPy
-
Input Source: Pre-recorded or live video feed
-
Detection Targets: Safety Helmet, Safety Vest
-
Output: Annotated frames with tracking and detection IDs
Key Features
-
PPE Detection: Identifies whether workers are wearing helmets and vests
-
Multi-object Tracking: Assigns unique IDs and tracks workers across frames
-
Optimized Frame Processing: Skips frames for real-time efficiency
-
High Accuracy: Trained on relevant safety gear data
-
Visual Overlays: Bounding boxes and clear labels for easy review
-
Manual Exit:
ESCkey to stop execution
Project Workflow
-
Load the custom YOLO12 model trained for helmet and vest detection.
-
Open the video file (e.g.,
helmet.mp4) and resize frames. -
Use
model.track()to detect and persistently track individuals. -
For each detected worker, draw bounding boxes and show labels (
Helmet,Vest, etc.). -
Continuously display output frames with PPE compliance status.
-
Exit by pressing
ESC.
Use Cases
-
Construction Site Monitoring: Automate detection of helmet and vest usage.
-
Workplace Compliance Auditing: Reduce manual labor in checking safety gear.
-
CCTV Enhancement: Integrate into existing surveillance systems.
-
Real-time Alerts: Extendable to alert systems for non-compliance.
Conclusion
SiteSafeTrack bridges the gap between AI and occupational safety. It empowers construction managers and safety officers with automated tools for real-time PPE compliance monitoring—reducing risks and ensuring regulations are upheld efficiently.

Reviews
There are no reviews yet.