UNICEF Innovation Fund Graduate: ekitabu
Using frontier technology & data to monitor child poverty in Iraq

Developing MERON: Using AI to detect how malnourished a child is through a single photo

The UNICEF Innovation Fund is proud to see portfolio member, Kimetrica  — graduate. They’ve come a long way – from numerous product iterations, to deep diving into understanding their ecosystem better, strengthening their business model, and gearing up to take their solution to market. They’re now ready to collaborate at a larger scale – as they find new pathways to work with partners, investors, and the open source community.

Reflections from the Kimetrica team:

Graduating from the UNICEF Innovation Fund, we’ve been able to develop a model that can predict children’s malnutrition at 60% accuracy – showcasing increasingly positive results when compared to traditional methods.  We collected over 4,000 images to train a model to measure malnutrition using machine learning and facial recognition. The team also created two fully connected neural networks to classify malnutrition status or predict a weight-for-height score, and developed a fully integrated system that enables transmission of photos from a mobile device or tablet to an API that returns a weight for height z score and malnutrition category.

Looking back…

There are two traditional methods of measuring moderate and severe acute malnutrition in children under five. One is mid-upper arm circumference (MUAC), and the other is weight for height (WFH). The problem with these methods, however, is that they require trained staff, bulky equipment, physical handling of the child, and can be time-consuming.


MERON uses artificial intelligence to detect a child’s level of malnutrition from a single photo.

Utilizing newly emerging facial recognition technologies and machine learning algorithms, we have developed a method for assessing malnutrition that is not dependent on enumerator assessment and aims to remove measurement, transcription and interpretation error. This method is not intended to be a replacement for MUAC/weight for height, but rather a complementary tool for assessing malnutrition.

Enumerators collecting data to train the model

Validating & Testing our solution: Fieldwork in Kenya to train the model

The next step in our model development was to test its application in detecting malnutrition in children under five. In order to train the model, anthropometric data and photographs were collected through field trials in four counties of Kenya. The data was collected in collaboration with the Ministry of Health (MoH) and UNICEF Kenya through the routine Standardized Monitoring and Assessment of Relief and Transitions (SMART) surveys.

We also wanted to understand how caregivers feel about using photography to detect malnutrition instead of traditional methods. Caregivers of children in all four countries confirmed that taking a photo is culturally acceptable. In three out of four counties (Isiolo, Turkana and Marsabit) caregivers actually preferred photographs compared to traditional methods of assessing malnutrition.

Next set of goals

In the next 12 months, we aim to improve the model further and develop tools so that it can be used more easily and effectively:

  • In order to continuously improve the model’s predictive capability, we need to collect additional images and further refine the model.
  • To ensure that the model performs well on different ethnicities, we will require images from different regions of the world.
  • We also plan to explore the inclusion of other types of data through mobile applications that can estimate the height of objects (e.g. Smart Measure and EasyMeasure).
  • We will develop a MERON App to enable easy photo capture by enumerators, caregivers or community health workers with minimal training required, and offer online and offline tools and quick and clear representation of diagnostic results.

In order to scale up the innovation and ensure that it achieves its full potential, we are raising additional funding to the tune of $300,000 and pursuing partnerships with aid agencies on the ground to collect additional images and anthropometric data.

We’ve had a lot of interest from the global community – MERON was presented at the Artificial Intelligence for Good Global Summit held in Geneva in May 2018 (Watch the interview here) and was featured in the Smithsonian, New Scientist, DailyMail and Deutsche Wella.

UNICEF Innovation Fund Graduate: ekitabu
Using frontier technology & data to monitor child poverty in Iraq