Introducing Facial Recognition

This is one typical scene in movies that Facial Recognition (FR) is put into action. For those using FaceBook, you will also recall that FaceBook automatically tagged the name of faces in the image that you upload. In physical world applications, FR is being used to verify an individual's identity in Immigration and making Payment.

FR is becoming a very common topic in the last few years, mainly due to the wide application in China for Law Enforcement, Payment, Customer Service, etc. There are both positive & negative outcomes that resulted from the usage of facial recognition

1. Chinese man caught by facial recognition at pop concert
2. UK police use of facial recognition technology a failure, says report
3. China’s ‘Big Brother’ surveillance technology isn’t nearly as all-seeing as the government wants you to think
4. Eyes set on the wanted in Penang
5. Lawmakers can't ignore Facial Recognition's Bias anymore

While the technology can achieved relatively high accuracy now as compared to few years back, the outcome of real application remains mixed. What I feel is the lack of holistic approach in deployment of FR where the focus is solely in technological aspect.

1. While the accuracy is high, what is the speed in matching and the hardware required to achieve this? Can I support my operation with this speed of return?
2. Does the placement of the cameras enable capturing of people's face or it is mainly seeing people looking away from the camera's view?
3. How's the environment going to interact to assist or limit the probability in getting faces?
4. How does Light Reflecting Markup affects the performance of FR? How about those Anti-FR approach to evade detection?

Applied FR on the field is both an Art and Science where one needs to take into consideration various factors that constitute to a good deployment. One need to consider things such as Infrastructural Cost, Hardware Cost, Customer's SOP, Human Behaviour, Response Time, Environmental Effect, etc. Putting all these together to achieve optimal performance given the situation on ground.

Just as each individual are an unique person, each deployment needs to be treated differently and no single deployment are exactly the same. Some required major rework on the approach while some just need minor tweaks to adapt to situation.

Regardless of your requirement, the considerations of having an optimised facial recognition solution remains largely similar. This article series on Facial Recognition aims to share with you what we have seen over the last 10 years on the field. Apart from technology and it's relevancy to each situation, we will consider some of the fundamentals that affects what you will be getting eventually.

Thank you for reading and stay tuned for the next article.