Facial recognition technology is becoming entrenched in our society; a part of everything from social media to law enforcement. How does the software determine identity with a photo, anyway?
Points of Recognition
Each face is unique, despite general structural similarities. Today’s facial recognition programs use a complex algorithm to compare an image of a face to other facial images in a database. These algorithms apply weighted values to each data point, or portion of the face, and ‘decide’ whether or not the photo is a match based on the final score.
Uses of Facial Recognition Technology
How is this technology being used? In a variety of ways, including law enforcement, entertainment, social media (Facebook) and even identity checks at airports. The British Airports Authority (BAA), for example, announced today that it will begin using facial recognition in Heathrow airport. In addition, many police departments across the United States are now using iPhones to check the identity of anyone suspected of a crime. This advanced software is also being used for entertainment, and has many other potential applications.
Earlier this week, I asked Viewdle, the company providing facial recognition for online games such as Third Eye, a few questions about their software. Marisol MacGregor, Head of Product Management at Viewdle, was gracious enough to provide details:
Decoded Science: How do you compensate for expression-changes when you evaluate for a match?
Viewdle: When we look at the face, our statistics for recognition are weighted towards the top 2/3 of the face, which is done to account for a change in expressions. We always need to be able to detect 2 open eyes and 40 px between the eyes. The more frames we have of you from different angles, the higher the precision-recall.
Decoded Science: Could you briefly explain how your integration of facial recognition into the online gaming market?
Viewdle: It’s more about face recognition being the first piece of the puzzle of a broader computer vision theme. People don’t yet understand that the processing power on phones is enabling new and amazing experiences within the phone’s camera, once digital elements are combined with your real world surroundings. Putting computer vision into games requires that the user interact with REAL things in front of them. We are seeing face recognition not so much as a “look down” experience, as when playing most mobile games, but as a “look up” experience.
Decoded Science: Do you foresee a more practical application for your product, such as the facial-recognition technology being implemented in law enforcement, or are you sticking with the entertainment/social media scene?
Viewdle: Practical. More people take personal photos than get scanned by these government things, so we think that our tech is very useful for the consumer market. i.e. Our faceprint solution for seamlessly sharing and organizing your media.
Invasion of Privacy?
Algorithms that can correctly identify people based on facial characteristics provide benefits for our society in various areas, from the relatively lightweight areas of entertainment and online gaming to the critical mission of terrorism prevention. In some cases, the use of software to chart and evaluate the human face may be an invasion of privacy. As with most technology, however, the software itself is not to blame for the potentially problematic applications or abuses by humans.
Decoding Science. One article at a time.