The face recognition algorithm invented by German researchers can achieve higher recognition accuracy without seeing the face in the photo.
Face recognition technology has caused people's concerns about privacy, and with the continuous advancement of related technologies, the privacy problems it creates will only become more and more serious. If the ability to identify a person's identity with just one photo creates a serious imbalance between the ruler and the ordinary citizen, the latest appearance of the "Faceless Recognition System (FRS)" completely destroys the only one. A little balance.
The faceless recognition system was invented by the Max Planck Institute in Saarbrücken, Germany, with the aim of identifying images through imperfections—blurred or blocked. The system uses a set of blocked and unblocked images to train a neural network to discover the similarities between the target's head and the body.
The test found that the accuracy of this algorithm is quite high. After seeing a photo with unblocked faces, it can identify the same person in the blocked photo with a 69.6% accuracy. And if the number of facial photos increases to 10, the accuracy will climb to 91.5%.
However, this technology also has its own limitations. If a person's face is blacked out, the accuracy rate will drop to 14.7%, but even this is more accurate than human judgment.
As for the privacy concerns it raises, the researchers responded like this:
“From a privacy perspective, the results presented by this technology do raise concerns. But similar undisclosed systems are likely to be put into use. We believe that the computer vision community has a responsibility to quantify and disseminate users online. Share the privacy issues with your photos."
ShenZhen Fahold Electronic Limited , https://www.fahold.com