Sistema Integrado de detecção de EPIS - SIDE

Abstract

This work presents the development of a computer vision system aimed at the automatic detection of Personal Protective Equipment (PPE) in workplace environments. The proposal seeks to integrate artificial intelligence technology with occupational safety, contributing to accident prevention and compliance with regulatory standards. The system uses the YOLOv8 model, a real-time object detection neural network trained to identify the main PPE items: helmet, safety glasses, face mask, and reflective vest. The dataset annotation and preparation were carried out using the Roboflow platform, which also facilitated the resizing, organization, and diversification of the images. For image capture, a camera connected to a Raspberry Pi was used, sending the data to a graphical interface developed in Python using the Streamlit framework. This interface allows real-time visualization of the detected equipment, making analysis easier for the user. The system was designed to be easy to implement and operate, offering a low-cost solution with potential application in various industrial sectors. Tests showed promising results regarding detection accuracy, even under different lighting conditions. The research highlights the potential of using computer vision and IoT to promote safer and more intelligently monitored workplaces. The combination of accessible hardware, intuitive software, and deep learning models proved effective in building systems that support occupational safety.