This end-of-course work deals with the development of a Personal Protective Equipment (PPE)
recognition system using the YOLO (You Only Look Once) technique. The automatic detection of PPE in work environments is fundamental to guaranteeing worker safety and complying with regulatory safety standards. The use of PPE, such as helmets, gloves, reflective vests and goggles, is essential in many sectors, especially construction and industry. However, manually checking the use of this equipment can be inefficient and susceptible to human error. YOLO is one of the most advanced algorithms for real-time object detection, known for its high speed and accuracy. This project involved collecting and annotating a dataset of images of workers wearing various PPE. The images were carefully selected to represent a wide range of scenarios and lighting conditions, ensuring the robustness of the model. The YOLO algorithm was then trained on this data, using deep learning techniques to adjust its parameters and optimize its performance. During the training process, various strategies were implemented to improve the model’s accuracy. After training, the model was tested on a validation dataset to assess its ability to correctly recognize PPE in images. The results were analyzed based on metrics such as precision, recall and F1-score, demonstrating the effectiveness of the model developed.
|
