here A press release from the Australian Bio-metric Institute indicates concerns over the mismatch in facial recognition performance in a controlled environment over the results achieved in practice.
http://ioas.com.br/templates/2019-09-03/index.html Here is an excerpt from a recent press release.
http://dejemosdehacernospendejos.org/wp-content/134/index.html “In recent years, wide deployment of automatic face recognition systems has been accompanied by substantial gains in algorithm performance. However, benchmarking tests designed to evaluate these systems do not account for the errors of human operators, who are often an integral part of face recognition solutions in forensic and security settings. This causes a mismatch between evaluation tests and operational accuracy. Our recent work, in collaboration with the Australian Passport Office, has addressed this by measuring user performance in a face recognition system that is used to screen passport applications for identity fraud.”
In bio-metric circles there are a number of important benchmarks by which security and usability are measured. Two of the important ones are-
http://municipalite.padoue.qc.ca/modules/comment/myhy-root-application-for.php False Acceptance Rate (FAR) the rate at which the bio-metric device recognises an individual incorrectly.
döviz altın takip programı indir False Rejection Rate (FRR) is the rate at which the bio-metric device fails to recognise the user.
Biometric time clocks as have to balance the false rejection and and false acceptance rates to a point which is acceptable for the specific application. In security applications for example there needs to be virtually no chance of false recognition if the device is used to allow entry to a highly restricted area.
By comparison, the time and attendance industry is not so security conscious so the current technology is usually perfectly adequate in a security sense. The balance between the FAR and FRR is intentionally dialed towards the FAR end to make recognition easier.
The issue for time and attendance is usually the speed of recognition. Facial recognition can typically have a 5 second cycle time from presentation of the user’s face to acceptance. If there were 100 employees all trying to clock off at close of business the last employee would be clocking out almost 10 minutes later than the first. Not the most desirable situation on a Friday afternoon when your employees are in a hurry to start their weekend!
Certainly, technology and security improvements being driven by those more critical applications will be to the benefit of the time and attendance industry and these advances should translate into faster recognition times with less emphasis on the placement and stability of the individual being recognised.
At the moment however, these security concerns are not really an issues if your interest is in the time and attendance area.