So, How Does Face Detection Actually Work? Part 1

With all our recent posts on face tracking and facial recognition, we thought it would be nice to write a little something on how face detection actually works. What are the mechanics of face detection? It is a pretty novel technology and most that use it on a regular basis, such as in a digital consumer point-and-shoot camera, likely don’t know how it works. Not that you need to know – as long as the camera is working, that’s all that really matters in the end. But for you more technical types, we found a pretty insightful article on how the details of face recognition come together to recognize a face.

What exactly is unique about the face that allows a camera to recognize a group of pixels as a face? Even more, it seemingly does it in realtime, which is quite astonishing if you stop and think about it. Cameras, of course, have become incredibly advanced in recent years, but even with all the technological features that come standard, there’s not a lot of room left for a chip equipped with extreme processing power. Some cameras now have software that can even tell when a subject is smiling so it can automatically take the picture.

Face recognition is due in large part to the work of Paul Viola and Michael Jones back in 2001. The two researchers created a framework for identifying arbitrary objects, which they would then refine to work with faces. In the industry, this was dubbed the Viola-Jones framework.

Face detection, says the original post, isn’t exact as a science. Software can be tripped up. And just as humans see faces and illusions in random things, so can computers. This is called Pareidolia, or the recognition of something significant, typically a face or human form within something that doesn’t naturally have it. Think the Virgin Mary in a potato chip, or the face of Mars, an image that appears to show a face etched in the ground of the Cydonia region of the planet Mars.

Viola and Jones’ method boasts a high rate of accuracy, and researchers have reported less than 1% false negative rate as well as a false positive rate under 40%, even when the most basic filter is applied (Viola-Jones works off a filtering of rectangular images). A full Viola-Jones framework incorporates up to 32 filters altogether.

We’ll cover some more of the details and mechanics in a second post – to be continued. In the meantime, have a look at the original article here:

http://www.techradar.com/news/software/applications/how-face-detection-works-703173

Related articles:

1. How Does Face Detection Actually Work? Part 2
2. A Deeper Understanding of Face Recognition
3. Ever Wonder How Facial Recognition Works?