How It Works
Our system runs on a local camera that continuously captures video. It leverages various APIs and models for detection:
- YOLOv8 (with Roboflow): Detects missing PPE such as helmets, gloves, and vests.
- MediaPipe: Implements real-time pose estimation to identify unsafe lifting techniques.
- Edge AI Processing: Uses YOLOv8 on-site for low-latency detection, reducing reliance on cloud processing.
- AWS Services: Manages cloud storage, data analysis, and compliance tracking. When an unsafe posture or missing safety gear is detected, the system automatically captures the current video frame and uploads it to an S3 bucket in JPEG format (with the image data stored as a NumPy array).
- Analytics Dashboard: Provides insights into safety trends and violations by aggregating data from AWS.
After inference, the trained models return the coordinates for the detected objects and annotate the live video stream to highlight which parts are missing or unsafe. This real-time feedback ensures that potential safety issues are flagged immediately.
How Site Sentinel Can Impact Society
By preventing workplace accidents and ensuring OSHA compliance, Site Sentinel contributes to safer work environments, reducing injuries, downtime, and costs. It empowers industries to proactively address safety concerns, improving overall productivity and worker well-being.