We all know those people who never forget a face; even infants display this facility to a refined degree. It seems to be a common characteristic of animals, but, up until recently, mimicking this ability using computer hardware has been extremely difficult and the purview of corporations and the government. Recent breakthroughs have even allowed this technology to be included on commercially-available laptops. So, how does this technology work?
Facial recognition software does much the same thing, relying on the geometry of the face to identify an individual. Different technologies excel under different conditions; while taking a picture of the subject using a conventional (video) camera may work acceptably under ideal circumstances, 3D imaging extrapolates those key characteristics to allow for improved identification from an angle. Surface texture analysis takes the key characteristics approach and applies it to the skin, enhancing the confidence of a match achieved via normal or 3D imaging. Like voice-recognition systems, many facial recognition systems also require training with photos taken under different conditions before they can reliably identify an individual. (See this RAND briefing for a more in-depth discussion of biometric identification as well as facial recognition.)
One long-term possibility for this technology is “autotagging”, automatic assignment of names to photos as they are taken. For those taking lots of photos or uploading into photo management software, such a feature could turn out to be very useful. A technical hurdle to be overcome is the processing power required for face recognition, but a limited reference set and continual increases in computing power may help to make this a reality sooner rather than later. So could a recent face recognition development that allows for identification of faces with minimal training.
In digital photography, face detection is a technique whereby the camera recognizes that there is a face in the frame and focuses on the subject’s face, adjusting both focus and exposure to optimize. While you may not see the usefulness of this feature under optimal circumstances, it comes into its own when shooting in low lighting or when in need of speed. It may not be a deal-breaker option, but it’s certainly worth putting in that “nice to have” category. While “face detection” and “facial recognition” may be used interchangeably, the first refers to confirming the presence of and focusing on a face while the second actually identifies the individual based on their face. Given the rapid pace of technological change, you’ll definitely want to stay posted for advances in both of these two fields!