Researchers from the University of Western Australia has developed a new innovation that could potentially revolutionize how security cameras Bunbury and across the country work, as well as the biometrics industry; a new, large scale 3D facial recognition system.
Facial recognition is becoming more and more commonplace, being used in an increasing amount of security cameras Bunbury and across the world thanks to more and more surveillance, security and IT industries using it. The technology is reliant on the ability of computer models to ascertain a person’s identity based on a scan of their face, to determine whether or not their legitimate.
Currently, facial recognition technology works on 2D scanning, having developed to the point of surpassing the recognition ability of humans, though there are some issues with the 2D model that the more advanced 3D model can handle. For example, 2D facial recognition has issues with changes to whatever it’s scanning, like changes in facial textures, expressions, posing and scale, issues that the 3D model does not have. The trade-off for the 3D model is that data for it can be difficult to gather.
2D facial data can be acquired by simply looking it up in the internet, whilst 3D facial data requires data to be physically collected from the real subjects themselves, which can limit their use.
The model, dubbed FR3DNet, was developed by a research team from the University of Western Australia’s Department of Computer Science and Software Engineering, and is the first of its kind. The team created the model by analysing the 3D scans of about 100,000 people, which amounted to 3.1 million scans.
The model was then trained by the team to learn the identities of a database of people then accurately match test faces to their respective identities.
According to the research team’s leader, Dr. Syed Zulqarnain Gilani, the model was a big leap forward for the 3D facial recognition field. He says that his hope is that this research will help improve facial recognition technology and software security across the world, in particular the devices that accesses networks and systems.
The FR3DNet model is currently available for any parties looking to do further research.