Advanced technology powered by Deep Learning
Aurora has used its own in-house face recognition technology for clients and partners since 2007. By the end of 2014 this had about 35 person years of development behind it and is amongst the most accurate systems available anywhere.
Aurora provides two main modalities of face recognition. First, for highly accurate one-to-one verification that is impervious to changes in lighting, clients have adopted our Infrared system with the optimised camera and light source that we have developed with our hardware partner, Perception Sensors & Instrumentation. Proven in numerous trials and in live implementations, the system is used widely in the air and construction industries amongst others. Second, our colour or natural light system which can be used with any off the shelf cameras.
Our systems can be utilised for face detection, feature point detection, classification, quality assessment, one-to-one matching, look-a-likes and best matches (identification).
In February 2014 Aurora achieved number one in the world status on the University of Massachusetts Labelled Faces in the Wild test site. You can download the full report here.
Since October 2014 our team of six PhDs, sourced from universities around the world, have been developing Deep Learning/Artificial Intelligence. The first task we set ourselves was to produce even better accuracy for our face recognition technologies. In August 2015 our on-going programme of development had delivered spectacular advances to our already accurate technology. There has been a reduction in error rates for our Infrared system of 65%, in our colour system of 44% and in our feature point detection an 85% improvement in precision. We anticipate further enhancements in the coming months.
See for yourself here; some of the comparisons with our competitors.
For partners that wish to integrate our technology to their applications we provide a range of Software Development Kits (SDKs).