A Deep Learning-Driven Computer Vision Framework For Real-Time Ppe Detection and Automated Safety Inspection

This study describes an automated system for verifying the use of Personal Protective Equipment (PPE) related to machine vision and the YOLOv8 model. The images were collected with a digital camera and analyzed through the OpenCV library. To evaluate performance, the paired T-test was applied, which showed a statistically significant reduction in verification time (p = 0.043). In practice, the average inspection time was reduced from 32.4 seconds under manual supervision to 16.8 seconds when the automated system was used. The normality of the data was verified by the Shapiro–Wilk test (p > 0.05) the system has a low operating cost of 7.49 USD per month. Economic analysis indicates a return on investment of 351.03% and a payback period of approximately 2.66 months. These results show that the proposed prototype can be applied in real industrial scenarios.