 My name is David Lilis. I'm a faculty member with the School of Computer Science and UCD and also a principal investigator for the Cedar Center for Applied Data Analytics. I've come to predict today to talk about the challenge of facial age estimation. So we look at people all the time and we try to figure out from looking at their faces what age they are. You have the example of the person in Tesco clicks the button to say you're definitely over 25 and you can buy that wine. And there's been a lot of work done over the last number of years in deep learning applied to this particular challenge. So you have lots of automated systems who look at photos of people and try to figure out what age they are. And it's kind of gone to the point now where if you compare the performance of these systems compared to the performance of humans doing the same task, they're pretty close. And that's interesting for certain challenges like filtering online content based on the age of the subject in a photo or things like that. Some of the interesting things that come out is we find that humans tend to overestimate the age of children, for example. And that comes across also in some of the automated services that we've surveyed. And it's also interesting that when you come to older subjects as well across the board the various different technologies that we've tested tend to underestimate the age of people who are in that older age bracket. So some nice insights coming out of that in terms of the performance of these systems at the moment.