Real Time Helmet and Mask Detection For Safety Critical Areas In The Workplace

The objective of the study is to detect the helmets and masks worn by workers, researchers, QA inspectors and any other individuals that work in a field/domain that requires a physical presence in potential safety critical areas and if it is found that the individuals are not wearing the proper protective gear, an alarm will be triggered. The paper provides a comprehensive examination of the field of intelligent access systems for safety critical areas. The purpose of this paper is to provide practitioners and researchers with a comprehensive understanding of the key challenges and opportunities associated with designing, implementing, and operating such similar systems. Five categories were identified for this examiner use case in the field of construction, demolition or environmentally hazardous workplaces: “head”, “helmet”, ‘'false mask”, “unmasked” and “masked”. A total of 1711 images of construction worker scenes were collected and enhanced to produce 4733 images. All images are annotated according to defined categories. Experiments split the training and validation datasets in a ratio of 9:1. This resulted in an average mean accuracy of 65.235% (graph).