Handwashing quality control using neural networks

Mārtiņš Ļuļļa, Aleksejs Rutkovskis, Atis Elsts, Maksims Ivanovs, Roberts Kadiķis

Research output: Contribution to conferenceAbstractpeer-review


Qualitative hand washing is one of the most efficient ways to prevent the spreading of infectious diseases. The aim of this study was to research handwashing quality control possibilities using neural networks. Neural networks and machine learning were applied to recognize handwashing movements and gestures. In order to teach neural network video files with handwashing episodes were collected and annotated using computer software. More than 1000 video files with handwashing episodes were collected in local hospitals. 700 videos were annotated - where each handwashing movement was labeled accordingly to World Health Organization recommendations. Annotated video files were used to train the convolutional neural network. As result, the neural network with an accuracy of 64% was obtained. The study shows that neural network accuracy was improving by increasing the number of annotated videos used to train the network. Therefore, we are planning to collect and annotate more video files to increase the accuracy and performance of the neural network in future studies. Our study shows that well trained neural network could be a beneficial tool in order to control and improve handwashing quality. Moreover, the neural network could be implemented in computer software, so the new type of handwashing quality control devices could be manufactured in the future.
Original languageEnglish
Publication statusPublished - 24 Mar 2021
EventRSU Research week 2021: Knowledge for Use in Practice - Rīga, Latvia
Duration: 24 Mar 202126 Mar 2021


ConferenceRSU Research week 2021: Knowledge for Use in Practice
Abbreviated titleRW2021
Internet address

Field of Science*

  • 3.3 Health sciences

Publication Type*

  • 3.4. Other publications in conference proceedings (including local)


Dive into the research topics of 'Handwashing quality control using neural networks'. Together they form a unique fingerprint.

Cite this