THERE鈥橲 a lot you can tell from a face 鈥� and now there鈥檚 one more thing: body mass index. Software that can predict your BMI from a snapshot of your face could turn a simple headshot into a revealing portrait of your build, and even your risk of certain diseases.
BMI is a standard health metric that鈥檚 equal to a person鈥檚 weight (in kilograms) divided by the square of their height (in metres). Someone with a BMI over 30 is acknowledged as obese and below 18.5 as underweight.
How we perceive someone鈥檚 weight based on images of their face is strongly correlated with actual body weight and associated health risks (Perception, ). Building on that knowledge, Guodong Guo and colleagues at West Virginia University in Morgantown have now produced an algorithm that can analyse a mugshot and predict that person鈥檚 BMI.
Advertisement
The software assesses seven weight-related components in a face image, including the ratios of cheekbone width to jaw width, face length to cheekbone width and the average distance between eyebrow and eye. They then ran the program across images of 14,500 faces of people with known BMIs. The predicted BMIs were mostly within two or three points of the person鈥檚 actual BMI (Image and Vision Computing, ).
That鈥檚 just a first attempt, Guo says. He says that tweaking the software to analyse more facial features should improve the results.
鈥淭his could be used in smart health applications, relating face images to BMI and associated health risks,鈥� Guo says. 鈥淥r on online dating sites, for instance, it could help you assess the BMI and state of health of people you might date.鈥�
聯The system could help you assess the BMI and state of health of people you might find on online dating sites聰
Ioanna Tzoulaki, an epidemiologist at Imperial College London, says such analyses could be useful if fat in the face can be correlated with patterns of disease, but she worries about discrimination. 鈥淲hat can this be used for except dating sites? People often have facial images in their CVs. Is it right to assign people into obesity groups from these?鈥�
