Automated Hand Hygiene Compliance Monitoring

Success of any commercial computer vision solution necessitates the use of right sensors and analytics over the data. RGB image sensors are proved to be efficient for deep learning methods to solve object detection, segmentation and classification problems. Very minimal works are reported in the literature that use other imaging sensors such as depth or infra-red. This work uses a deep learning method over depth sensor data to solve one of the needed solutions of hospitals, battery plants, and restaurants, the Hand Hygiene Compliance Monitoring. Our technical contributions are of three-folds: (1) we evaluate the working of transfer learning using RGB data pre-training and depth data re-training (2) we design a generic strategy for the application as existing solutions in the market are performed under controlled environment (3) finally we compare our deep learning solution performance with a conventional machine learning solution.