Researchers from the University of Western Australia has recently developed a new innovation for security systems Bunbury and across the country and the world, with a brand new system capable of handling large-scale 3D facial recognition operations.
This newly developed model could see use in the future from any organisation or government agency in need of for more accurate 3D and could also see widespread use, improving security systems Bunbury and across the world, with the potential of making personal passwords entirely obsolete.
Current technology operates on 2D facial recognition of photographs, and while the general consensus is that it has surpassed what human recognition is capable of, it still has several issues and shortcomings that the 3D model can take care of. Unlike their 2D cousins, 3D models can potentially handle changes in facial texture, expressions, poses and scale, though the issue is that the data is rather tough to gather.
2D facial data also has the detail that it can be easily acquired by looking it up on the internet, whilst 3D facial data requires actual collection from physically present subjects, which limits its use but also stops it from being fooled by pictures from the internet.
The research team behind the development, from the UWA Department of Computer Science and Software Engineering created the model, dubbed the FR3DNet, via analysing 3.1 million 3D scans of their sample, which amounted to about 100,000 people. They then trained the model to familiarize itself with the dataset of ‘known’ people and then match one of the test faces to the identities found in the database.
According to leader of the team, Dr. Syed Zulqarnain Gilani, the 3D model was a huge leap forward for the field of 3D facial recognition. He says that, with off-the-shelf 3D now becoming cheaper and more affordable, the future of pure 3D facial recognition systems doesn’t seem as far fetched as before.
Their research, Dr. Gilani says, shows that recognition on 3D scans is superior to that of 2D scans, allowing for scanning in any pose, wearing accessories like glasses or face masks, as well as different expressions and the model can recognize the person instantly.
The research paper was published in Computer Vision and Pattern Recognition, and the 3D Facial Recognition model (FR3DNet), is available for anyone looking to do further research on it.