In February 2020, Kimetrica presented its facial recognition software at a workshop on Artificial Intelligence and Children in Cape Town, South Africa.
The workshop, hosted by UNICEF and the Government of Finland, is part of a global initiative that brings together governments, businesses, civil society, and children to develop policy guidelines on how Artificial Intelligence (AI) can impact children, both positively and negatively.
At the workshop, Kimetrica’s Chief Research Officer, Dr. Claire Simon, presented Kimetrica’s Methods for Extremely Rapid Observation of Nutritional Status (MERON) technology. MERON, which uses model ‘trainable’ neural networks to extract facial features, involves taking a photograph of a child’s face on a cellphone or a tablet. The child’s facial features, combined with information on gender, age and ethnicity, can predict the weight-for-height (WFH) Z-score in children under five, thus identifying malnutrition and the severity of it.
As a result of the conference Kimetrica identified several firms who are interested in partnerships to expand MERON's reach. Kimetrica is currently exploring various partner arrangements, including seeking additional funding to improve MERON's accuracy and reach.
"MERON is one positive example of how the power of AI can be used to protect vulnerable children across the globe. It is our hope that as MERON uses machine learning to improve, it will more rapidly identify malnutrition from a diverse set of facial features, allowing us to save tens of thousands of lives," Dr.Simon said.
Data collection using the MERON App
Pilot research suggests that, when collecting data in the field, MERON outperforms the traditional Middle-Upper Arm Circumference (MUAC) method of measuring nutrition levels in children. MERON is also less invasive and cumbersome to use, as well as more cost effective than MUAC. Kimetrica has completed its proof of concept data collection in Northern Kenya, and is currently seeking grant funding to collect more images to improve the accuracy of the model..
Since the photographs of children are automatically deleted once the image dimensions have been uploaded to the cloud, the model poses less security risks to children, offering protection against invasion of privacy or identity theft.
Kimetrica is certain that MERON’s performance will greatly expand as its access to photographs of diverse populations across the globe grows.
Once MERON achieves high-quality classification ability, it will offer the following benefits:
- An increase in the accuracy of collecting data on malnutrition;
- Less expensive enumerator training;
- Use of inconspicuous measurement tools;
- Ease of use under challenging field conditions;
- A less invasive method to measure malnutrition in young children.
Kimetrica provides software, research, survey, modeling and impact simulation services for evidence-based decision-making and learning. Kimetrica works with governments and non-profit organizations to increase the impact and efficiency of their social investments, enhance accountability, manage critical risks, and build donor or taxpayer confidence. Kimetrica has run projects in over 50 countries around the world and continues to expand its reach. Kimetrica has offices in Washington DC, Broomfield, Colorado, Nairobi and Addis Ababa. Find out more at www.kimetrica.com.