-Title: An Improved YOLOv9-Based Object Detection with Attention Mechanism for Personal Protective Equipment Monitoring in Industrial Environments
-Conference: European Conference on Artificial Intelligence (ECAI 2025), Oct 26–31, 2025, Santiago de Compostela, Spain
-Authors: Geunho Lee, Jieun Lee , Jongpil Jeong
-DOI:
-Conference Link: https://ecai2025.org/

Abstract: In real-world industrial environments, numerous potential hazards exist and accidents can occur at any time. To ensure worker safety, it is essential that personal protective equipment is worn by all employees. To overcome these limitations, there has been a growing emphasis on the utilisation of automated detection technology.
The present study proposes a method for accurately detecting personal protective equipment on workers by combining the Convolutional Block Attention Module with the YOLOv9 model, a model widely used in object detection. Although YOLOv9 offers high detection speed and accuracy, it is susceptible to false positives and false negatives under various conditions in real-world industrial sites. To compensate for this, the Convolutional Block Attention Module is incorporated into the model.
In this paper, model training and evaluation were conducted using a dataset of manually annotated personal protective equipment from images of workers collected in various industrial environments. The experimental findings demonstrate that the YOLOv9 model when combined with the Convolutional Block Attention Module exhibits superior performance in terms of precision and recall in comparison to the conventional YOLOv9 model.

-Status: Submitted(2025/05/07)