It may not exactly be back-of-the-head recognition, or male-pattern-baldness biometrics, but Google Photos appears to have developed a way to suggest who a person is in a photo of them that shows the back of their head, rather than their face, according to Android Authority.
The app has long used facial recognition for tagging photos with the names of those in them, but the new feature shows accuracy estimated by Android Authority at around 80 to 85 percent at identifying an individual in shots where they are facing away from the camera. The method of doing so may take signals other than the appearance of the person’s head, such as their clothes, similarities to other photos where their face is visible, or image meta-data.
To whatever extent that biometrics are playing a role in the new method of identification used for tagging by Google Photos, it appears to be something entirely different from the challenging task of matching face biometrics for faces that are heavily occluded or shown at acute angles.
The photos shown and described in the article do not include the subject’s eyes, nose or moth at all, and would previously not have been tagged by the app. Now, a notice reading “face available to add” suggests the identity of the subject, and tags the photo if it is confirmed by the account operator.
Google Photos has also reportedly gotten better at identifying face partially occluded by masks, no doubt with the benefit of the millions of images uploaded during the pandemic that it could use for training.
Article: Google Photos now suggesting names when no face visible