UvA computer scientists present smile database

Over four hundred people, old and young, from all over the world, contributed to the richest and most extensive smile database ever collected. Computer scientists from the University of Amsterdam's (UvA) Faculty of Science recorded the smiles of hundreds of visitors to the NEMO science centre in Amsterdam. This makes it possible to understand precisely what happens with your face when you smile, and how this affects your apparent age.

At NEMO, 481 subjects participated in the research of computer scientists Theo Gevers and his colleagues Albert Ali Salah and Hamdi Dibeklioglu. From each subject, a posed smile and a genuine smile was captured on video. They also had to pose as angry, happy, sad, surprised and scared. Certain characteristics, such as the speed with which the corners of the mouth move upwards, were analyzed. These features can be used by computers for automatically estimating the age of a person, for recognizing emotions and for human behavior analysis. The research was conducted as part of the project Science Live, sponsored by the Netherlands Organisation for Scientific Research (NWO) and the Royal Netherlands Academy of Arts and Sciences (KNAW). You can see samples at www.uva-nemo.org.

Software that knows how old you are
The researchers also asked the subjects to look at images of other subjects. They had to estimate how old these people were and how attractive they found these people. They were also asked to estimate some traits: Is this a helpful person, or is he perhaps in love?

With all the collected data, the researchers could develop software that can estimate the age of a subject. This software takes into account whether someone is happy, sad or angry, and adjusts the age estimates accordingly. The software appears to be slightly better in estimating ages than people: we are on the average on target with an error margin of seven years, whereas the computer has a margin of six years. Together with color scientist Dr. Marcel Lucassen, the group has also shown that females look younger when they smile, but only if they are over forty. If they are under forty, they should look neutral if they want to come across young.

Theo Gevers has been contributing to research on computer analysis of faces for several years now. He was awarded an NWO Innovational Research Incentives Scheme VICI grant in 2007 for his pioneering work in computer vision.

Other content in...