Masks Impact on Face Recognition Algorithms

Feb, 2021 - By SMI

Face masks help limit the spread of germs, it is one of the best protection against the spread and limit of COVID-19, unintended effect: disturbing facial recognition algorithms.

A study in the U.S. National Institute and Technology (NIST) found wearing a face masks that covers mouth and nose causes the error in widely used facial recognition algorithms to nail between 5% and 50%. Whereas, blue masks were less likely to cause errors than black masks and the more nose portion covered by the mask, the tougher the algorithms found it to identify the face.

The new government study reveals that how the face masks adoption might end up making stronger facial recognition algorithms and the algorithms deal with face masks than they do about how companies are already hard at work building algorithms that can adapt to new situations. This error caused by mask wearing is not too surprising. These algorithms are generally trained to identify based on facial geometry although the mask hides a huge portion, which means wearing a mask can reduce the accuracy of algorithms. According to the study, the best of 89 facial algorithms were tested which had error rate between 5-50% in matching digitally applied face mask with photos of the same person without a mask. Analyzing the experiments, it was noticed that shape and color changes the performance of algorithm. Facial recognition was more accurate in people wearing blue or round mask than people wearing black mask.

Meanwhile, this would seem as a good news for those who are worried about their privacy. This technology would be enthralling even when it works and when it doesn’t, as the accuracy and privacy issues are resolved in the algorithm.