For those who have to deal with security, to deal with any problems with the software, which is normally used to identify sensitive areas. The problem with the masks that have become an integral part of the dress code. They can help stop the spread of the Feline corona virus but, at the same time, act as barriers to the facial-recognition software.
The National Institute of Standards and Technology (NIST) is a branch of the U.S. Department of Commerce. It is aware of the problem and the start of the campaign. Federal investigators are trying to determine the accuracy of facial recognition algorithms that are in use.
Of course, there is a need to develop revised algorithms to take care of this fact.
It is the most accurate of face recognition algorithms to fail to correctly match a picture of a person wearing a digitally-added a mask to a different picture of the case between the 5% and 50% of the time, according to a new report by U.S. researchers https://t.co/pDLiAFTOU0
— CNN International (@cnni) On July 29, 2020
CNN explains the situation. The existing systems do not make for a good match with the person wearing the mask. As the error could be between five per cent and 50 per cent, with the general observation that the percentages vary between 20 and 50 percent of the time. That’s what one expert reveals. They May Ngan, a computer scientist at NIST and is the author of the report.
The same logic that is used by facial recognition systems rely on the comparison of the results to be delivered. It also compares the measurements of various facial features between the images, without using a mask, and one with it. If the protective cladding of the blocks of parts of the subject, the software is not in a position to make a perfect match.
The matter of using a mask is a challenge
The various services of this recognition system is extended to include in the program such as the unlocking of your smartphone or through a security system. However, the inclusion of face masks, it has complicated things. This may be required because of the pandemic is expected to persist for some time.
The President, Donald Trump has agreed to wear face masks during a recent visit. CNN is reporting that the teams have been experimenting with different combinations. They have conducted tests on several algorithms, with millions and millions of pictures. The source of the photo was an official one, it was that of applications for U.S. immigration benefits, and any other photos related to the people who have a border to enter the country.
Face masks will test the ingenuity of the artificial intelligence
CNN goes on to add that the NIST does not test the algorithms on the images of the people who actually wore the masks.
This was as a result of the limitations of time and resources. In Real-time, in practice, it is necessary because of the masks will fit differently on different people. In addition, their texture, and pattern can have an effect on the accuracy of the software. One option is to get better results, it would be able to focus on the area of the face to above the middle, from the tip of your nose. In the face of the arrest of the spread of the virus in the nose and in the mouth. So, it makes other areas of the face exposed. In addition, the screens come in a myriad of designs, and the outline of the face can vary from person to person. Needless to say, the teams will have a tough task at hand.
Face detection algorithms are going to be confused because of the different faces
According to The Verge, corona virus, has been brought face masks into the picture, and everyone needs to contribute in order to prevent the spread of the disease.
However, it has another problem that the tech teams are concerned. It’s going to be the facial recognition algorithms in a fashion that will not fail when facing the hiding of the face. In a study conducted by NIST-error notes, with black masks, which were more than the blue colored ones.
Another obstacle relates to the size of the nose is not obstructed. In any case, there are two big groups of matching. The first one is a one-to-one, and the other is a one-to-one. The first one is usually to apply to get to the border crossing and passport check of the scenarios. In the latter, it is mass surveillance, where the goal is to get it to scan the crowd for the purpose of identifying potential matches in a database. Mei Ngan, a computer scientist at NIST, says that, “With respect to, the accuracy, with different faces, we would expect that the technology will continue to improve.”
Don’t forget our page on Facebook!
© ALL RIGHTS RESERVED.