Does Facebook’s face recognition levels its accuracy to human brain? This has become the most trending issue in the world of technology and social media. The trend begins when Facebook acquired Face.com in 2012 which is specializing on facial recognition technology. Now, Facebook moves forward into a more comprehensive technology. Its research project, DeepFace is developed with more advance functionality and is believed to be as accurate as the human intellect. Yaniv Taigman is one of the developers of DeepFace technology who was also the founder of Face.com. The new facial recognition technology can recognize faces by tracking various pictures across the web. Report says that while humans can recognize faces on 97.53% accuracy, Facebook’s facial recognition can recognize at 97.25% accuracy rate. It’s not bad at all. In fact, it’s almost perfect!
Although DeepFace is just a research project, a lot of people found it helpful as it could be a very powerful tool for transparency on the cyber space. The project was developed by Facebook Al research team in Menlo Park, California. The new software is supported with a neural network, a single program that is capable of copying approximation on the composition of neurons. It works by analyzing data, like in the human face; it goes on a wide search of data through various patterns to achieve an accurate result. Basically, the software is composed of clustered neurons (nine layers deep), along with the examination of connections between the neurons and millions of face images across the web.
After tough examination process, data gathered (images) are feed into the system transitory to the synapses producing distinct mark below the nine layers of neurons processing several conditions like for example, is the face having thick brows? If the answer is yes, it follows one synapse. Else, it follows another route. This goes the process of face recognition.
DeepFace can identify faces on 2 different photos with different lighting angles. However, there was an instance where Facebook tries to impress the public but the result was not good as expected when the person being recognized on the photo commented that it wasn’t him. However, an explanation was given. The software can really do tough research across the web comparing a particular photo to other existing millions of photos and links back to a particular profile assuming that the profile contains photos of the owner. Generally, Facebook gives recognition every time a user uploads photos on his profile and prompts the user with tags if the system found the same matches.
At the moment, a lot of people are still not convinced with Facebook’s face recognition accuracy. However, with continuous development, everything is possible. DeepFace is still on its research stage which means, it will move further on various studies and tests until perfection is achieved. Perhaps, it will only be a matter of time.