Vision-Based Hand Hygiene Monitoring in Hospitals

Hand hygiene has been shown to be an effective intervention to reduce transmission and infections in many studies. This project focuses on interpreting visual clinical data for hand hygiene monitoring. We propose two distinct deep learning approaches to detect hand hygiene action on manually collected and labeled data. Specifically, we investigate a fixed-window and a pose-based hand detector using Convolutional Neural Network (CNN). We show both approaches are able to achieve high accuracy and outperform our baseline model using linear Support Vector Machine (SVM) classifier.