Facebook has learned to recognize faces in the crowd with a precision of 97, 25%

Facebook has developed an algorithm called DeepFace, which allows you to identify the person in the crowd with an accuracy of 97, 25%, which almost corresponds to the ability of the average man (97, 53%), writes TechCrunch.

To overcome the limitations of conventional programs for facial recognition, developers have found a way to build Facebook 3D-face model on the picture. These models can then be rotated to ensure recognition of a face imprinted at different angles. In the past, the same attempt to recognize a person could easily be unsuccessful if the person is just slightly tilted his head to the other side.

To the algorithm DeepFace they must be trained on a large number of persons. In the current version it can identify up to 4 thousand people on the basis of a database of more than 4 million individual images. Theoretically, this database can be expanded to apply to most social network Facebook, it would be helpful if Facebook wants to automate the process of identifying all of your contacts and recognize faces in photographs without the need to manually mark users.

So far, this project is designed only as a research paper published last week, and a group of authors is going to present the results of the Conference on Computer Vision and Pattern Recognition in Columbus, Ohio, in June. Nevertheless, the project has great potential for future applications, as in the Facebook, as well as in the field of artificial intelligence.

Source: habrahabr.ru/post/216281/