Deep learning for the detection of good hygienic practices control measures for food handlers

Abstract

Hygiene habits and behaviors of food handlers are critical to ensuring food safety during production. Good Hygiene Practices (GHP) require food handlers to follow strict hygiene standards, including proper handwashing procedures and the use of personal protective equipment (PPE) such as hairnets, masks, gloves, protective clothing, pants, and work shoes to prevent cross-contamination in food factories. However, manual inspections in large-scale operations are often inefficient and impractical. To address this, this study proposes a deep learningbased system that leverages computer vision and Convolutional Neural Networks (CNNs) to automatically detect hand hygiene and PPE compliance before workers enter the production area. This ensures adherence to GHP standards, enhances food safety, improves inspection efficiency, and reduces costs. A dataset of 20,222 entries, comprising 12 handwashing actions from 8 angles and color-coded PPE conditions, was used to train and evaluate 10 models: YOLOv8, YOLOv7, YOLOv6, YOLOv5, ResNet, Dense-Net, MobileNetv2, EfficientNetv2, VGG, and Vision Transformer (ViT). YOLOv6 and YOLOv8 achieved the highest accuracy (0.999) for handwashing recognition, while Dense-Net achieved the highest accuracy (0.956) for PPE detection. This system offers an efficient and automated solution for monitoring hygiene habits, helping to prevent cross-contamination and ensure food safety within factory environments.