Empowering Hand Hygiene monitoring through Video Annotation
Introduction
In the realm of public health, effective handwashing plays a crucial role in preventing the transmission of infectious diseases. BlueBerryTech took on a contracted project centered on video annotation for handwashing improvement, utilizing the Computer Vision Annotation Tool (CVAT). This case study explores a significant dataset recorded at Pauls Stradins Clinical University Hospital, featuring 3185 handwashing episodes among medical staff. The annotations adhere to World Health Organization (WHO) guidelines, creating a robust dataset for training machine learning models and investigating real-time handwashing Objectives:
- Annotate handwashing videos using CVAT.
- Develop a labeled dataset for training machine learning models.
- Evaluate CVAT’s efficacy in accurately annotating handwashing actions.
- Investigate the potential for real-time monitoring of handwashing compliance.
Methodology
BlueBerryTech initiated the project by organizing tasks and uploading video datasets to CVAT. Handwashing movements were meticulously labeled, aligning with WHO guidelines. Handwashing movements were meticulously labeled, following a classification system in line with WHO guidelines. The movements encompassed:
- Hand washing movement — Palm to palm
- Hand washing movement — Palm over dorsum, fingers interlaced
- Hand washing movement — Palm to palm, fingers interlaced
- Hand washing movement — Backs of fingers to opposing palm, fingers interlocked
- Hand washing movement — Rotational rubbing of the thumb
- Hand washing movement — Fingertips to palm
- Turning off the faucet with a paper towel
- Other hand-washing movement
Following the annotation process, a rigorous quality assurance procedure was implemented, and the data was stored in YOLO format for seamless integration with machine learning models.
Key Benefits of Annotation with BlueBerryTech
Pattern Recognition and Model Training
Annotation empowered AI models to recognize specific handwashing patterns, automating labeling and enhancing efficiency.
Temporal Context
for In-depth Analysis: Video annotation introduced temporal context, enabling AI models to understand hand movements over time for in-depth analysis and control of disease spread.
Cost-Effective and Efficient Data Generation
BlueBerryTech highlighted the cost-effectiveness of obtaining more data points from a single video, streamlining the process for improved efficiency.
Applications
The handwashing videos serve a dual purpose – as a foundation for training machine learning classifiers and as a platform to assess real-world handwashing quality in healthcare settings. This contributes to improved hygiene practices and automated handwashing movement recognition.
Conclusion
BlueBerryTech’s successful completion of the project, utilizing CVAT for annotating handwashing videos, demonstrates a commitment to advancing public health through technological innovation. The labeled dataset not only serves as a resource for training machine learning models but also holds the potential to revolutionize real-time hand hygiene monitoring, fostering a safer and healthier environment in medical facilities and beyond.