Artificial intelligence against child malnutrition
What if weight and height of a child could be determined from a photo…
How can we measure the weight and height of a child? The first idea that comes to our mind is usually weighing scales and height gauges used at the pediatrician’s clinic. Unfortunately reality turns out to be much more complicated in several developing countries. The Human Measurement Project aims to develop a tool which will be able to extract weight and height of a human being from two pictures: a frontal and profile picture.
In order to set up this tool, an algorithm will be developed from a very large photo database. Data collection will take place in many different countries and has already started in Cambodia thanks to a partnership with UNICEF and RACHA (Reproductive and Child Health Alliance). Magali Cattin, a master student in Bioengineering and Human Measurement Project intern, had the opportunity to join the Cambodian team for one week. Her stay was brief but intense. When she arrived there, nearly 1,300 pictures had already been taken. Based on these samples, Magali was able to identify challenges met by the photographers. Pictures should be taken in a specific manner in order to facilitate the development on the algorithm, but there is often a mismatch between theory and field. Furthermore, the image collection must be conducted under very strict ethics regulation. At the end of her stay, Magali met the data collectors and trained the photographers for the upcoming field collections.
The Cambodian data collection will go on for a few more weeks and other collections are planned during the next months in different countries thanks to local partners. The project is part of the EssentialTech programme of CODEV and gathers many institutions at EPFL: Professor Jean-Philippe Thiran’s signal processing laboratory (LTS5), professor Robert West’s Data Science Lab (dlab) as well the Swiss Data Science Center (SDSC). Besides, external organisations are taking part to the partnership such as Centre Hospitalier Universitaire Vaudois (CHUV) and Terre des Hommes NGO.
Swiss Data Science Center