Despite the great popularity and interest in facial recognition tools – and similar systems – has grown in recent years, that its use has become popular in the unlocking of various devices (among others), and that at our disposal we find a wide range of alternatives, the truth is that there are not so many open source options; that is to say, the software of this type of open source.
A reality that has led a group of researchers from Carnegie Mellon University in Pennsylvania to develop and launch one -whose name is quite revealing-: OpenFace. Curiously, it has done so only two days after Panasonic presented its (WV-ASF900).
The possibilities of Open Face
Thus and as others do, Open Face –based, by the way, on FaceNet, the Google research project- is capable of capturing and identifying faces in real time; to achieve it, yes, you need at least ten photos of the person in question.
To start using it, you must identify each subject, completing an informative section; and install a series of requirements so that the program learns to use all its features.
In addition, the system, which has been tested with over 6,000 images personalities such as Emily Deschanel, Amy Adams, América Ferrera, Bradley Cooper, Eva Longoria and Ben Stiller, among others; it takes “a few seconds” to find the match and is 87% accurate.
If you want download it, you will find the OpenFace source code on GitHub, where you will also find more information about it and you have the option to ask your questions about it.
In any case and regardless of what has been said, although he still has a long way to go, it seems that the software it is going in the right direction. “Despite the fact that the public data sets with which we are operating are still smaller than those of the private industry, the precision is very high and surpasses all other open source implementations of facial recognition”, comment its creators.
Via | The Next Web
Photo | Pixabay (edited)
In Genbeta | Facebook’s facial recognition scales globally and tags our photos automatically