Implementation of a convolutional network for detection of PPE in automotive repair services.

Resumo

This study explores the application of computer vision to occupational safety in automotive services, specifically focusing on the real-time detection of Personal Protective Equipment (PPE). The automotive service context presents unique challenges in managing the proper use of PPE, such as safety glasses, gloves, and appropriate uniforms, to ensure the safety of those involved in service execution.

Identifying and monitoring the correct usage of PPE is crucial for maintaining work quality and preventing workplace accidents. The research aims to develop and implement a methodology tailored to the specific needs of automotive services, with a focus on repairing windshields, headlights, and mirrors. A detection model based on the YOLOv8 convolutional network is proposed for its simplicity and potential for integration into a portable platform.

The study's objective is to achieve satisfactory precision in PPE usage detection. Additionally, it seeks to optimize the efficiency of automotive services, avoiding interruptions and delays in service delivery due to workplace accidents. The research builds on prior studies that demonstrated the effectiveness of computer vision in PPE detection in industrial and construction settings. The proposed approach includes creating a dataset and training a convolutional network with approximately 2000 images from security cameras.

The research significantly enhances both computer vision and occupational safety in automotive services, reducing the number of non-compliances found in internal inspections and improving the quality and safety of the workplace. Moreover, it addresses challenges related to managing image banks and the complexity of detecting PPE. The study also examines successful experiences in other sectors that used convolutional networks for real-time PPE detection.