SafelineAI Moggers 24-D

This image is currently processing. Please be patient...

Inspiration

Industrial accidents caused by the lack of proper use of personal protective equipment (PPE) are a significant concern. With thousands of lives lost each year, we were inspired to leverage AI to create a solution that ensures workers are consistently using their safety gear in hazardous environments. Our goal is to prevent accidents before they happen and improve workplace safety across industries.

What it does

SafeLine AI uses real-time video feeds to detect whether workers are wearing required PPE, such as hardhats, safety vests, and masks. The system tracks and monitors compliance, sending this data to a dashboard that evaluates metrics like the percentage of safety equipment usage and overall risk. This allows for real-time feedback and alerts when workers are not in compliance, helping to avoid potential accidents.

How we built it

We used YOLOv8 for object detection, specifically trained to identify PPE like hardhats, masks, and safety vests. For tracking, we integrated DeepSORT to monitor individual workers throughout their shifts. The system sends detection data to a MySQL database hosted on Google Cloud, which is then visualized through a Grafana dashboard to display safety percentages and risk metrics. The architecture relies on real-time processing of video feeds from factory lines to ensure accurate and timely data collection.

Challenges we ran into

One of the main challenges was ensuring the AI model accurately detected PPE in various lighting conditions and angles. Integrating the DeepSORT tracking with our detection system to ensure workers were individually tracked over time presented some technical difficulties. Additionally, managing the real-time data flow between the detection system, the database, and the dashboard was a challenge, especially when handling large volumes of video data.

Accomplishments that we're proud of

We successfully created a real-time, scalable solution capable of accurately detecting PPE usage across various industrial environments. The integration of object detection with tracking and data analysis has allowed us to create a powerful safety monitoring tool. We’re particularly proud of how we’ve combined AI with data visualization to not only track safety compliance but also provide meaningful insights to improve workplace safety.

What we learned

Through this project, we learned a lot about the intricacies of real-time video processing, object detection, and tracking. We also deepened our understanding of how critical data management is when dealing with large volumes of real-time data. Additionally, the project taught us how to balance accuracy and performance in AI systems deployed in industrial environments.

What's next for SafeLine AI

Next, we aim to improve the accuracy of our PPE detection model by training it with more diverse datasets. We also plan to expand the system’s capabilities to detect other safety violations, such as improper use of machinery. Further, we hope to integrate predictive analytics to help industries forecast potential safety issues before they occur. Finally, we envision scaling SafeLine AI to become a comprehensive safety management platform for various industries.