Facial Recognition Technology explained... in under 100 words
An increasingly interconnected and complex world presents consumers with new possibilities in the customized delivery of goods, services, and experiences. This trend shows no signs of slowing, but presents public and security leaders with the challenge of ensuring stability without infringing the rights of citizens and businesses.
Artificial Intelligence has already provided solutions to many of industry and society’s modern problems, notably through the development of facial recognition technology. This type of technology uses the digital imaging of a user’s physical appearance (usually the face) to create a profile that is stored in a private database. The system remains dormant until a future input matches the stored image, which then triggers a pre-determined response.
It’s 7:00 and you’ve arrived at your office earlier than the receptionist who normally lets you into the building. Without thinking, you remove and scan your biometric ID over a panel. A camera activates, takes your digital photo and the office door swings open.
Your office’s facial recognition technology just executed a ‘quality check’ – assessing similarities between the image previously stored on your company ID, and the photo just taken. The system also analysed dozens of your unique physical dimensions and determined that your appearance matches the benchmark in its databank. Therefore, it has allowed you to enter.
How does it really work?
Facial recognition technology is designed to mirror the way people actually view and remember new acquaintances, for example by their ruby red hair. In the same way, a facial recognition platform creates a digital ‘memory’ of the new subject and stores it as a profile. It can then allocate, update and store additional data to the profile every time there is a match between a received input and the benchmark profile, or image.
Unlike human memory, however, the technology leverages sophisticated algorithms to create a numerical basis for each unique profile. A particular platform may use an image to generate specific dimensions based on the distance between an individual’s nose and chin, or between their two eyes, for example. The more dimensions a system receives, the greater the profile’s uniqueness but also the more efficient the system becomes at identifying exact matches.
Is it scalable and will it grow?
The market for facial recognition technology, with applications in security and social media, is poised to double in size between 2017 and 2022, up to an estimated US$ 8 billion per year. Increases in digital imaging resolution, combined with computer and processing power growth, ensure that facial recognition technology will expand in scope and efficiency. For example, Facebook’s own image identification algorithm can already identify matches with up to 97 per cent efficiency.
The scope of facial recognition does not end with consumer identification alone; it also extends to recognising products throughout the manufacturing process. Manufacturers can leverage the technology to automate detection of counterfeit products and defects in assembly lines. Known as ‘machine vision’, the concept is primed for piloting in stock-keeping and replenishment in both warehouses and retail outlets. One of the best examples of this is the Amazon Go store.
What are the technology’s limitations?
Despite growing attention to its broad applications across industry and marketplace, there is still concern surrounding its ethical applications and privacy restrictions. With cameras constantly filming and capturing images, citizens may feel powerless to prevent companies and individuals from capturing, storing, and analyzing their likeness, actions and preferences in an inaccessible database.
Similarly, some facial recognition platforms perform poorly in low light areas and are less adept at distinguishing between people of different ethnicities. This presents additional ethical and efficiency-related challenges as it limits system performance and potentially places individuals at risk of misidentification.
How can industry leverage the technology?
Industry and consumer excitement for facial recognition technology will doubtless continue to propel its maturity and refinement. However, for the time being, best practices suggest leveraging facial recognition as a supplemental function, rather than displacing existing processes until the technology can develop more completely.
This symbiotic approach focuses on capturing gains from the technology while minimising exposure to known deficiencies. Prime examples of this approach might be the way facial surveillance recognition supports traditional identity checks in airports or the way Facebook confirms whether a user would like to ‘tag’ a contact in a photo.