TechTheDay

It’s a fact that facial recognition technology is nothing compared to the human ability to distinguish faces. Even the continually progressing Facebook face recognition tech is still far from becoming perfect. Nevertheless, there have been exciting advancements in this technology.

Interestingly, in a paper published in March,  an artificial intelligence system referred to as “DeepFace” was reported to have closed the gap between human and machine face verification accuracy. The paper was presented at the Conference on Computer Vision and Pattern Recognition (CVPR) 2014. It was reported that DeepFace is capable of correctly matching faces 97.25% of the time.  This number is very close to the 97.53% accuracy rating achieved by humans in the same study.

This is indeed a very promising development in the field of facial recognition. Before DeepFace, all facial recognition software have accuracy problems when it comes to scanning faces that are not in full frontal views. With DeepFace tech, recognizing side view faces can now be more accurate. Even better, this technology may become a part of Facebook’s facial recognition system.

Here are more details you may want to know about this exciting technology.

What is DeepFace

DeepFace is a system that thoroughly analyzes facial features based on over 120 million different parameters. It is designed to emulate the way the human brain processes visual data. DeepFace analyzes huge amounts of visual information related to a specific visage and tries to draw out connections.

Relying on a technology called “deep learning,” DeepFace more thoroughly evaluates facial features by first coming up with three-dimensional models of the face being examined before rendering judgment.

Who Developed It?

DeepFace is a new technology developed by Facebook researchers. The technology is proprietary. Facebook executives, however, did not specifically cite DeepFace as something that will be exclusively used (if it is indeed going to be used) on Facebook.

How Does It Work?

DeepFace was developed after examining patterns in more than 4.4 million tagged faces on Facebook. The R&D team behind it meticulously picked out possible patterns and links that can be used in more accurately identifying faces.

Instead of simply evaluating faces as flat surfaces, a 3D model is being generated in black and white. Fiducial points and contours are determined to come up with a 3D model that can be viewed from different angles to extrapolate new views of a single face from different directions. As such, side view faces can now be associated with faces that are in full front view. Reciprocally, front views can now be associated with faces in side views.

When Will It Come to Facebook?

Unfortunately, it is uncertain whether or not Facebook intends to bring this to the social media service. According to Facebook spokesperson Lydia Chan, DeepFace is considered as part of “theoretical research” at Facebook. There are no fixed plans on bringing it to the social media service yet. Of course, this also means that the current Facebook face recognition system does not employ DeepFace tech yet.

It would be understandable for Facebook not to bring this advanced facial recognition system to Facebook for free. As mentioned earlier, the system involves millions of parameters being analyzed. Hardware upgrades may be needed to be able to run it properly, in its full potential. Also, something this advanced and close to human accuracy could be offered as a premium service. It has the potential of becoming a Facebook revenue driver.

Do you need it?

At present, the most common use of facial recognition technology on Facebook is the automatic tagging of Facebook users present in photos. The overwhelming majority of Facebook users are not really concerned about accurate facial recognition on social media. While DeepFace can bring significant improvements in the identification of people in photos posted on Facebook, it is not something that can be considered essential. Many may not even be aware of this news or are likely going to ignore this piece of new information about their favorite social media site.

Accurate facial recognition is best used by authorities who need to investigate people to solve crimes. It is more useful to those who are authorized to track someone for valid reasons.

Issues

DeepFace is indubitably a step forward in technology but it is not without its issues. For some, this new facial recognition tech can be an advantage but for others, DeepFace can be a nightmare. There are Facebook users who want some extent of privacy so they would prefer not getting tagged or identified on group photos taken by their friends or taken at big social events. Many may find it inconvenient or irksome if DeepFace starts demonstrating its “creepy” accuracy on social media.

DeepFace is certainly a great feat in facial recognition technology. However, it’s uncertain if this new technology can be practically useful to typical Facebook users. While it can lead to better automatic tagging on Facebook, some may complain that it is somewhat intrusive and irritating. Also, its more complicated information analysis may require hardware and software upgrades to cope with the higher level of processing needed.