Site Sentinel

Inspiration

Purdue University is constantly evolving, with ongoing construction projects ranging from new apartment buildings to academic renovations. Seeing these frequent construction sites, we became concerned about potential workplace accidents. Since safety hazards can arise suddenly, we wanted to create a solution that proactively prevents accidents before they happen.

What it does

Site Sentinel is an AI-powered Industrial Safety Monitoring System that enhances workplace safety through:

PPE compliance detection using Roboflow to ensure workers wear helmets, gloves, vests, and goggles.
Posture monitoring with MediaPipe, identifying unsafe lifting techniques in real time.
Cloud-based analytics dashboard for tracking safety trends, generating reports, and ensuring OSHA compliance.

How we built it

YOLOv8, trained with Roboflow, was used for detecting missing PPE such as helmets, gloves, and vests.
MediaPipe was implemented for real-time pose estimation, identifying unsafe lifting techniques.
Edge AI processing with YOLOv8 enables low-latency detection on-site, minimizing cloud dependency.
AWS services manage cloud storage, data analysis, and compliance tracking.
A cloud-based analytics dashboard provides insights into safety trends and violations.

Challenges we ran into

Every technology we used—Roboflow, MediaPipe, YOLO, and some of AWS services—was new to us. Learning them within a short timeframe was challenging.
Integrating local detection functions with AWS services required precise handling of outputs to ensure stable cloud communication.
Balancing real-time processing with accuracy and efficiency required extensive testing and optimization.

Accomplishments that we're proud of

Successfully integrating AI-powered PPE detection and pose estimation into a real-time safety system.
Deploying AI for low-latency processing, making our system efficient and scalable.
Creating a cloud-based analytics dashboard that provides valuable safety insights for managers.
Overcoming steep learning curves with new technologies and building a working prototype within a hackathon timeframe.

What we learned

How to train and deploy object detection models with Roboflow.
Real-time pose estimation using MediaPipe to track worker movements.
The complexities of integrating AI with cloud-based analytics.
Best practices for AWS services, ensuring stable cloud communication.
The importance of real-time processing for workplace safety applications.

What's next for Site Sentinel

Featuring hazard detection to include falling objects, gas leaks, and slippery surfaces for a more comprehensive safety solution.
Improving the efficiency and accuracy of YOLOv8 PPE detection and MediaPipe posture analysis to reduce false positives and enhance reliability.
Developing a mobile app for instant safety alerts, compliance tracking, and on-the-go monitoring.
Enhancing dashboard functionalities with AI-driven predictive safety insights, allowing managers to anticipate and prevent potential risks.
Exploring partnerships with industry leaders to bring Site Sentinel into real-world industrial and construction environments.