My initial research into automatic facial recognition (AFR) has thrown up a long list of links. This blog is going to be really useful in helping me to select the ones I have the most to learn from.
There is a mass of information online partly because the research in AFR began in the 1960s, and partly because governments appear very enthusiastic to support its development. The main aim is to make AFR more efficient and effective – to bring its success rate up to 100%.
I had no idea there were so many different forms of AFR. Basically the software needs to measure the face and then compare these measurements to a database of faces. There are many ways to capture a face, such as: plotting points and making measurements, using infrared light, using 3D scanning, analysing skin texture.
The databases themselves throw up a multitude of problems. The early databases consist almost entirely of white males (reflecting, I guess, the associates of the computer science researchers), with later databases featuring women and people who aren’t white caucasian. See this article in the Guardian.
This link below lists databases from universities that are are available for use. http://www.face-rec.org/databases/ – a fascinating insight into the mechanics of AFR research – I also find myself questioning the ethics of how these databases are compiled in the first instance. One dataset consists of women (and presumably young women and girls) who are there doing YouTube make up tutorials. Once online they have no consent with where their image goes or how it is used.
