The Digifort Facial Recognition system is based on the SAFR platform, the world’s premier Facial Recognition system for live video. It identifies camera unaware individuals in motion, occluded, in various head positions up to 90 degrees from the camera, under different lighting conditions and with changes in facial features such as ageing, varying hair styles, facial hair, glasses, face paint, or makeup. The platform identiﬁes individuals with a 99.8% accuracy rate.
Digifort Facial Recognition features include:
99.8% accurate facial recognition in real time (with age recognition)
0.048% Wild Faces FNMR* compared with eighty other systems according to NIST Wild Faces (National Institute of Standards and Technology).
Instant recognition (under 200ms) operating 6x channels with 5x faces in each field of view, simultaneously on the same server.
Gender, age and expression independent.
Scaleable across multiple servers with central management of a watch list (authorized and unauthorized users) for site management and access.
Capture faces starting from 50 x 50 pixels.
Management of the registration of people in the SAFR database through a Smartphone App or through the SAFR Desktop Application.
Web interface monitoring with alerts in real time to emails, SMS and mobile.
GDPR compliant with personal data and encryption.
*The NIST Wild Faces FNMR (False Non Match Rate) score of 0.048 found that the SAFR algorithm correctly recognized a camera unaware individual from an imperfect image in 95.2% of cases while perfectly differentiating a population of 10,000 people.
How the Facial Recognition system works.
The Facial Recognition system uses an image or video of a person’s face, ideally taken from the front, to form a recognised person database. An algorithm converts the image or video image into a numeric template. This cannot be converted back to an image for security reasons. Every numeric template is different, even if it is an image of the same person. However, those from the same person are more similar than templates from different people. Images of the same person can be linked on the database to build up an increasingly robust template profile of that person.
When the Facial Recognition system is operating in real-time, templates of people’s faces are taken and compared to those in the database. The technology identiﬁes individuals by matching the numeric template of their face, with all the templates saved in a database, in a matter of seconds.