The RecFaces updated biometric solution Id-Guard has become more accurate in recognizing faces, despite the pandemic and the mask regime
The COVID-19 pandemic has made a difference in people's lives. This includes the announcement by the state authorities of a lockdown with the mandatory wearing of medical masks in transport, shops and public institutions, and the expansion of the usage of remote work and distance learning, and even the introduction of a curfew and the closure of borders by countries.
At first, as more people began to wear medical masks due to the pandemic, the prevailing view was that their use posed a serious problem for facial recognition. It seemed quite obvious, since the algorithms designed for such analysis showed less recognition accuracy when part of the face was hidden. After all, for the reliable operation of most facial biometrics recognition algorithms, it is required for the face to be as open as possible.
This fact was confirmed by researches. For example, the US National Institute of Standards and Technology (NIST) tested 89 commercial facial recognition algorithms and found out that in 5 to 50% of cases there are errors when comparing photos of the same person's face with and without digitally added masks.
However, the developers proved the fallacy of this judgment, quickly creating improved algorithms that no longer require maximum face openness.
For example, after updating one of the main products of RecFaces the Id-Guard solution for biometric identification based on the image of a face in a video stream, it now only needs almost 50% of the face of a masked person for the most accurate recognition.
The company said that the used algorithm takes into account the points, lines and angles of the face, the location and distance between the eyes, etc., so half of the face above the mask is quite enough for identification. And it no longer plays any role, what type of mask a person uses.
After integrating an improved version of the algorithm into the RecFaces’solution, the product became more accurate and faster to recognize faces. Moreover, by increasing the recognition angles to 90 degrees vertically and horizontally, Id-Guard identifies the number of people moving in the flow by 20 % better than in the previous version of the solution. Now the system is able to recognize a person's face in the frame from different angles, regardless of the rotation and position of the head.
By increasing the accuracy of the recognition system, RecFaces specialists have also expanded the functionality of the product. Now it «helps» to make sure that the person in the frame is «alive», notifying the administrator in case of falsification. When an unauthorized person appears in the frame, an alert will also follow, which eliminates the passage of two people through the check point at the same time.
Now you can also add desktop widgets to the internal corporate website, such as the video stream from the camera, information on detected people, tracks, etc., which makes it easier to work with Id-Guard, making it more convenient.
The updated solution aroused the interest of national and foreign companies that send requests for testing and using Id-Guard.
Like all RecFaces software products, the Id-Guard solution integrates seamlessly with the video management systems, ACS, etc., available on the site, and the presence of the API makes it a universal face recognition tool for any video surveillance system installed on the site.
For example, Id-Guard is fully integrated with the Milestone Video Management System (VMS), and all the features of the solution are available in the Milestone XProtect interface, Bosch Video Management System, and Pelco VideoXpert.