Posted on 21 April 2016
Dr David Robertson, from York’s Department of Psychology, used PsychoMorph to blend together each previous companion’s images to form an ‘average’ female and male face.
By averaging together the faces of 14 different actors and 27 different actresses to have played the role, Dr Robertson mimicked the process that the brain is thought to perform when learning to recognise new people.
This is a different use of the PsychoMorph software, which ordinarily creates an average image based on different instances of the same person.
In a recent study led by Dr Robertson, the use of a person’s face average, rather than an individual photo of them, meant the accuracy of a smartphone face authentication system was significantly improved, sometimes to perfect levels.
This finding comes at a particularly important time in the development of biometric authentication systems. For example, new software uses face recognition to secure and unlock data in laptops, while some companies are introducing ‘selfie pay’, allowing users to pay for goods online using their face to authenticate the payment.
Dr Robertson suggests that the incorporation of face averages into such systems would undoubtedly bring benefits to their face recognition accuracy.
He said: “There have been lots of problems with automatic face recognition software in the past, but here we have developed a method of improving the system simply by using a face average rather than an individual photo. I would urge companies to look closely at the benefits this could bring.”
“As a fan of Dr Who, I really enjoy making these face averages and it is interesting that the companion averages show a certain resemblance to current actors. Initial responses suggest that the female average looks somewhat like Oscar nominated actress Felicity Jones, while the male average is thought to bear a resemblance to Thor actor Chris Hemsworth. “
The creation of the ‘average’ companion follows on from the Doctor Who face average that was created from each of the actors to have played the Time Lord.
Wednesday 23 October 2024
Wednesday 23 October 2024
Sunday 20 October 2024
Thursday 17 October 2024
Wednesday 16 October 2024