Can you guess a person’s name by looking at their face?

Do you look at someone you don’t know and think “that person looks like a Ulysses?” New software has shown that it’s possible to take an educated guess at a person’s name with just one look at their face (New Scientist, 6 April 2013).

Researcher Andrew Gallagher of Cornell University in Ithaca, New York, says that parents’ choices for their children’s names are influenced by factors such as gender, ethnicity, friends, family, and popular names or influential people at the time of the birth. Names are important. In related research, the team cites that “implicit egotism” is often used to decide upon a child’s name—that is the tendency to gravitate toward people, places and things that resemble the self. This, the team states, is why a disproportionate number of people choose a partner with a similar name to their own, such as Eric and Erica, and why more people with surnames beginning with Cali- live in California, and why people with the name Baker are more likely to be bakers.

In his team’s research, tagged photographs were used to build a database of named faces. The team programmed a computer to recognize the contributing factors (beyond gender and age). The team tested their system by using faces belonging to people with the top 100 names in America—48 male, 48 female, and 2 neutral—which cover approximately 20% of the population. With 640 training images per name, the team calculated the predication accuracy and Mean Average Precision (MAP).

The computer was able to guess the correct name 4.17% of the time which was significantly better that the random guess performance of 1% accuracy. Researchers say the “good” result was due to the fact that names are not randomly distributed across people (i.e. that people do not choose their babies’ names at random). They say that it is due to many correlations existing between given names and various facial features, such as skin colour, masculinity or femininity, size of face, age, and possibly even nameless attributes.

With improvements in accuracy, the software could be used to automatically tag people from their picture, called facial processing by modelling the relation between first names and faces, or building a matched face pyramid using multi-feature Support Vector Machine (SVM). Researchers maintain that “such a system, even if imperfect, could have a broad range of applications in security (e.g. finding fake person identities in a database) and biometrics (e.g. inferring the gender, age and ethnicity by guessing likely names). The research will be presented at the Computer Vision and Pattern Recognition conference in Portland, Oregon, in June 2013.

Other scientists say that there may be commercial applications for this software, but it will be hard to improve the accuracy because of the sheer volume of different names in the world—not to mention the number of names that are “conceived” each year!

H. Chen, A. Gallagher, B. Girod, "What's in a Name: First Names as Facial Attributes," IEEE Conference on Computer Vision and Pattern Recognition 2013.

http://chenlab.ece.cornell.edu/people/Andy/research/research.html

MARTINA NICOLLS is an international aid and development consultant, and the author of:- Similar But Different in the Animal Kingdom (2017), The Shortness of Life: A Mongolian Lament (2015), Liberia’s Deadest Ends (2012), Bardot’s Comet (2011), Kashmir on a Knife-Edge (2010) and The Sudan Curse (2009).