Millions of people are starting to wear masks in public after a U.S. Centers for Disease Control and Prevention (CDC) recommendation last month that social distance scares to help slow the spread of the new coronavirus could not be sustained,media reported. In today’s social media-driven world, it’s clear that masked selfies and photos will soon flood on popular platforms such as Instagram and Facebook.
This is already happening and has become an important issue for facial recognition platforms.
However, when wearing the mask properly, it covers most areas of the wearer’s face, including the nose, mouth and chin. In response, facial recognition companies are trying to retrain their algorithms to make them more dependent on other attributes, and they are reportedly using photos from public social media accounts to do so.
Media CNET said thousands of “masked selfies” had been found in public data sets. The data on GitHub, for example, included more than 1,200 samples, all classified by people wearing medical masks, not wearing masks or non-medical (DIY) masks. It also cites more than 5,000 photos from China a month ago of people wearing masks.
Artificial intelligence start-up Workaround, which provides 1,200 samples, said they collected it from the public account on Instagram. Wafaa Arbash, the company’s CEO, told CNET that they found the images by searching tags associated with the masks. However, these users are not required to be allowed to appear in the database.
In response, a Facebook spokesman who owns Instagram told the media that it does not allow third parties to collect or use photos uploaded by users in this way without their consent. “We are continuing to investigate this matter,” the spokesman added.