Facial recognition. Three common pitfalls, and how to fix them
Facial recognition systems are becoming more and more popular. More and more companies of different industries are using them. Key indicators such as recognition accuracy, along with a decrease in operating costs, are becoming the main parameters by which, the effectiveness of these systems is measured.
Mikhaylo Pavlyuk, CCO, 3DiVi Inc. explains what challenges operating experience brings in facial recognition systems. Including those that are not obvious at the start of the project. Fixing them can make your processes more efficient, and your business more successful.
Mixed photo quality in the database makes a promised recognition rate unreachable.
We do not understand what quality photos we have in the comparison database, and we cannot adequately build decision thresholds “yes / no”. Thresholds are set either “by eye”, or based on some general recommendations.
The result is not optimal operation of the recognition system, a multiple increase in errors of the first and second kind. Additional cost for processing erroneous comparisons.
The cost per mismatch incident is highly dependent on the business case.
If we talk about ACS, then this is the time spent by the security service on re-identification in case the correct person is not missed, or the time and effort spent on localizing the consequences of the penetration of a “stranger”.
If we take a more complex case – identification during a banking transaction, then this is the price of a possible loss of a client if we do not recognize him, or the admission of a stranger to the user’s bank accounts.
Let’s consider a business case of bank application login through biometrics.
According to statistics, the average number of logins per user is 19 to 25 times a month. Take, for example, a bank with a base of 1 million customers. We receive at least 19 million identifications per month. The best NIST algorithms for the highest quality Visa Photos dataset show FNMR = 0.0006 at FMR = 0.00001
In practice, having a dataset that is not quality checked, we will get at best FNMR = 0.003 with FMR = 0.00001.
This is more than 45,000 “extra” incidents per month. If each incident costs us an average of $1 (which is a grossly underestimate), then we get from 450,000$ per year of “extra” costs**. This amount is quite commensurate with the annual cost of owning a biometric system.
Using the photo quality control algorithm, we build a procedure for regularly checking the dataset. We remove and replace low-quality photos.
Cost savings commensurate with the cost of ownership of a biometric service, or 0.002$ per 1 transaction.
Lost customers due to the lack of feedback about face image quality during biometric enrolment.
Customer acquisition costs a lot of money. And sometimes the client receives a denial of service due to an uncontrolled environment and the inability to quickly prompt what needs to be changed to obtain the face image with enough quality. Example, we are trying to go through the authorization procedure in a dark room, forgetting to remove the mask. The system is wrong both in matching and liveliness evaluation.
The result is a loss of customer loyalty or loss of a customer.
The cost of attracting one new client to the bank for lending services is at least $20. The bank’s profit on average from one client is at least $90. In total, each client is $110 in costs and lost profits.
Consider the business case of remote issuance of loans to customers by the Bank. Suppose the Bank issues about 10,000 loans per month using biometric identification. On average, at least 9% of customers try to pass biometric identification in inappropriate conditions. In total, we have about 900 attempts at risk. Assuming that we finally lose 30% of leads, then for the year our lost profit will be more than $ 350,000.
With the help of the quality control algorithm built into the mobile application, we inform the client in about the problems of the environment and the necessary actions on his part, which significantly increases the chance of achieving the desired result.
Increasing profits from transactions related to remote identification by at least 10%.
Face recognition systems (ACS, Safe City) lack mechanism for monitoring the stability of the quality of incoming face images.
- There is no quality control of the face images received from the cameras and used for the recognition procedure. In this case, if camera starts sending low quality images, facial recognition systems that lacks an adequate control mechanism let the entire process related to face recognition to deteriorate.
- Each facial recognition camera should be specially configured for the best quality. The lack of objective quality control data for settings in terms of suitability for further face recognition procedure leads to the need for additional work in the future.
A significant deterioration of the quality of face images entails an increase in the number of missed detections and a decrease in the probability of correctly recognizing people from the list. Approximately 2% of cameras installed inside buildings, and more than 7% installed outside buildings, during the year, under the influence of external factors, begin to significantly degrade the quality of the initial data for recognition. Reduced system performance increases the cost of installing additional end equipment and processing capacity.
In one of projects, one of the camera lenses was harmed by coating with hairspray. This led to a 15% decrease in the number of face detections, which resulted in a decrease in the recognition of the control group by 8%. At the same time, it was problematic to notice changes in quality on the computer screen with a naked eye.
Using the quality control algorithm, you can organize statistical evaluation of quality of images for each camera and set notification if the quality deteriorates beyond a threshold.
Reduced operating costs to ensure guaranteed quality of installed system throughout all term of operation.
About the author
Mikhaylo Pavlyuk, CCO of 3DiVi Inc., is a sales expert and biometrics solutions specialist with over two decades of experience in the financial market and retail. Pavlyuk has helped hundreds of businesses in 10+ countries develop efficient and secure face biometric systems with solid architecture, fast scaling and integration.
DISCLAIMER: Biometric Update’s Industry Insights are submitted content. The views expressed in this post are that of the author, and don’t necessarily reflect the views of Biometric Update.