By Abhijit Shanbhag
Facial recognition is becoming an important tool for many applications involving people and cameras. As we continue to exit out of this global pandemic, facial recognition and more broadly facial analytics will continue to build in importance, even as the associated technical challenges and required capabilities grow.
There have been various significant advances in leveraging some of the latest applications in AI technology for facial recognition. According to research published by the Centre for Strategic and International Studies (CSIS), facial recognition systems have nearly absolute precision in ideal conditions, reaching a 99.97 per cent recognition accuracy level. It has to be noted, however, that conditions were far from ideal in many scenarios, and with the uncontrolled environment of CCTV on the streets, the accuracy is commonly lower even with more advanced algorithmic techniques.
We commonly divide the facial recognition capability into two different types: FR-1 and FR-N. FR-1 is to identify in a binary way if a person is the hypothesized one (mainly for authentication) and FR-N is to detect if the person is from a database of a certain number of faces. With FR-N, the size of the database can typically range from a few dozen as with a small enterprise to many millions of people in a large city.
The Unique Identification Authority of India (UIDAI) is currently in the process of incorporating FR-1 to bring in an additional layer of authentication for financial services requiring Aadhaar. Aadhar enabled Payment System (AePS) transactions will be authenticated using the person’s biometrics which may include FR-1. Various leading banks are in different phases of rollout of FR-1 to carry out such authentication. Similarly the Indian telecom operators will also be using FR-1 with an on-spot photo, which will then be verified with eKYC, prior to the SIM being activated.
Given the current pandemic, the inherent contactless nature of FR makes it a highly desirable alternative to fingerprint scanners and even RFID cards for office access and attendance monitoring. Many enterprises in India are now beginning to leverage FR-N for their attendance management leveraging their installed CCTVs at the entrance. They can also easily include additional analytics to detect tailgating, mask compliance, social distance compliance and have a full-featured AI-powered solution leveraging their installed CCTVs.
Within the transportation segment, multiple airports in India will soon leverage FR-1 at immigration checkpoints, and also at automated immigration clearance gates which will be unmanned. Increasingly some of the airports globally are also beginning to deploy video AI-powered technology to identify consumer sentiments and experience within the airports, and we expect this to also be considered within Indian airports in future. The taxi and ride-sharing industry has also adopted facial recognition to bring an added layer of verification and security to rides.
Some of the Indian telecom carriers have plans to deploy some of FR services at homes, SOHOs and condos or apartment complexes. The key use of facial recognition is on intrusion systems that detect if someone enters a home when an intrusion alarm is left armed. Automatic access is given to invited visitors who have RSVPed prior. When installed in personal properties like housing complexes, offices, it also comes with the feature of enabling individual blacklists and whitelists for different locations and entry points, thus preventing any unidentified individual from entering the premise and eliminating trespassing or intrusion. This can easily be expanded to prevent people that are not compliant with mask usage from entering the complex.
Some of the retailers and even fast food restaurants are beginning to leverage FR at an early stage. FR allows retailers to capture what shoppers are looking at in physical shops, obtain their sentiments allowing actionable reports to on-ground retail operators. Similarly various customer experience metrics from entry to pick-up and exit can be obtained, leveraging FR for fast food restaurants.
Facial recognition technology is also hugely beneficial in ensuring democratic voting, with the ability to detect fraudulent activities like booth capturing, false voting, identity theft and so on. The machine will only let you cast your vote after it has identified you as the correct voter.
Facial recognition can change the way we live our lives. However various technical challenges will continue to be worked through, such as the issue of face masks and ageing. Current developments will improve the application of facial recognition technology and iron out any issues along the way.