Facial recognition technology has improved dramatically over the past few years. In the latest round of testing conducted by the National Institute of Standards and Technology (NIST) in March 2020, showed that the best face identification algorithm had an error rate of just 0.08%. In 2014, the leading algorithm had an error rate of 4.1%.
In 2018 Face Recognition Vendors Tests (FRVT) carried out by NIST, NEC’s facial recognition technology ranked No. 1 for the fifth time, achieving an error rate of just 0.5% – a significant improvement from 2014 and the jump again in 2020 is significant.
In the 2018 tests, NIST found that more than 30 algorithms had achieved accuracies surpassing the best performance achieved in 2014. The continuous research and development of facial recognition technology is contributing to these improvements in overall accuracy and helping to create more use cases for facial recognition technology in a time when we are moving towards a more contactless society in light of the COVID-19 pandemic.
According to research published in April 2020 by the Centre for Strategic and International Studies (CSIS), facial recognition systems have nearly absolute precision in ideal conditions, reaching a 99.97% recognition accuracy level. However, perfect conditions are hardly achievable in daily operations, and algorithms face various factors affecting their accuracy.
How accurate is facial recognition technology under real-world conditions?
Accuracy scores of 99.97% are possible under ideal conditions where there is consistency in lighting and positioning and where the facial features of the subjects are clear and unobscured. In fact, under those conditions, facial recognition accuracy scores are comparable to the best results of iris scanners. The trouble is those conditions are challenging to achieve under real-world conditions.
There are a number of factors that can impact the accuracy of facial recognition technology under real-world conditions. These can include:
Ageing is a natural process that has a notable impact on the accuracy of facial recognition systems. The skin texture changes shift the facial highlights.
While the best algorithms achieve impressive results even as subjects age, changes such as wrinkles and face shape alterations may make a person less recognisable for a facial recognition algorithm, especially for the oldest among us.
Partial covering of the face can happen due to wearing a medical mask, sunglasses, spectacles, earrings, and scarves. It may also occur due to hair, moustache, or a beard. This is an issue that is becoming more prominent due to the COVID-19 pandemic, with many governments enforcing the mandatory use of face masks, especially on public transport and in airports.
Facial coverings deteriorate the accuracy of all facial recognition algorithms. However NEC has already launched a facial recognition system that identifies people even when they are wearing masks, adapting to a new normal where face coverings have become a key form of protection against the spread of the coronavirus.
NEC was already working on a system to meet the needs of allergy sufferers who wear masks – a common practice in Japan – when the COVID-19 pandemic prompted it to accelerate development. You can read more about this technology in a recent Reuters post.
Low-resolution images often derive from surveillance cameras. Individuals captured in such photos are often a crucial part of an investigation but are challenging to identify compared to a high-resolution database.
A low resolution can have a negative impact on accuracy rates. To achieve reliable face recognition of a video image, NEC developed feature point extraction technology that enables enhanced face recognition accuracy to a level where an individual can be identified with high precision from within a group, even if their face is partially hidden, or the image is taken from different angles. NEC’s face recognition technology also uses deep learning technologies for face matching to increase accuracy to a level where an individual can be identified by a low-resolution face image captured by a distant camera.
The way the light and shadow fall on a human face always affects appearance. Thus, photos taken in varying lighting conditions may be potentially challenging for a facial recognition algorithm.
Even though illumination distortion increases error rates of modern algorithms, there are ways of coping with its effect to receive accurate facial recognition results.
How will facial recognition be used in 2021 and beyond?
Facial recognition is still a technology that courts controversy. Headlines around the world often detail the perceived misuse of facial recognition technology and the possible infringement on privacy and civil rights.
However, for all the negative press around the use of facial recognition technologies, there are countless case studies showcasing the positive outcomes that can be achieved by using facial recognition technology. As we move through 2021 and beyond, we can expect facial recognition to become more immersive in our day to day lives as we battle to overcome the COVID-19 pandemic around the world.
From international travel to access and authentication, businesses are looking at ways to keep their staff and customers safe. Reducing the number of touchpoints an individual must make on a day to day basis is one way that businesses are helping to tackle the spread of COVID-19 and potentially other viruses that can be passed by close contact.
Read more about the uses of facial recognition technology in 2021 and beyond in our recent post.
NEC and facial recognition
Currently, NEC’s face recognition is independently recognised as the fastest and most accurate face recognition software in the market. NEC’s NeoFace® Watch was ranked top in Face In Video Evaluation (FIVE) of Non-Cooperative Subjects 2017, Face Recognition Vendor Test (FRVT) 2013, Multiple-Biometric Evaluation (MBE) 2010 and Multiple Biometric Grand Challenge (MBGC) 2009, independent tests conducted by the internationally renowned US National Institute of Standards & Technology (NIST) and peer-reviewed by the scientific community.
The tests position NEC’s face recognition software as the most accurate face recognition software even with low-quality images. Independent tests also demonstrate NEC provides the fastest matching capability most resistant to variants in angle, age and race.
This doesn’t mean we are resting on our laurels – R&D continues to improve both the speed and accuracy of our facial recognition software while keeping security at the forefront of our developments.
The world in which we live is definitely changing and facial recognition will become an integral part of our day to day actions if it isn’t already. In fact, if you are reading this article on your mobile device, it is highly likely that you unlocked that phone with the aid of in-built facial recognition.
- Why should we care about facial recognition in 2021?
- Which biometric authentication method is most secure?
- 5 most common uses of facial recognition
- Seamless travel – how facial recognition is revolutionising travel and security
- Identification and payments at a distance: touchless and contactless experiences in a post-COVID-19 world