Using Computer Vision to Detect Personal Protective Equipment

In a guest post edited by the Roboflow team, data scientist Jaco Lau discusses his development of a computer vision application to detect Personal Protective Equipment (PPE) on construction sites, highlighting the critical role PPE plays in preventing injuries and fatalities. Despite the high incidence of safety infractions due to improper PPE use, Lau created a model leveraging computer vision to identify safety violations by using object detection to recognize whether PPE was worn correctly or not. Lau initially faced challenges in gathering suitable images but overcame them by manually selecting relevant data and using Roboflow to annotate and preprocess images, which included resizing them to a uniform dimension for faster training. The application prioritizes recall to ensure it captures all safety violations, albeit with some false positives. The final model was integrated into a Streamlit application for deployment, showcasing the ease of using Roboflow for image processing and Streamlit for application deployment, significantly streamlining the development process.