German computer scientists proved that there is really nowhere one can hide with a discovery of a technology that can recognize a person's face even in darkness. Researchers at the Institute of Anthropomatics & Robotics and Karlsruhe Institute of Technology announced a new method that relies on a person's thermal signature to determine their identity. Details are revealed in the new study by Saquib Sarfraz and Rainer Stiefelhagen available through arXiv.org.
The new technology is widely seen as impressive because the use of thermal signature as main determinant of facial features and identity is very problematic. The signature can always change especially once a person changes his facial expression, which happens often. In addition, thermal signatures also vary according to the way it reflects light and the temperature of the person's skin and environment, the Daily Mail noted.
The researchers were able to address the thermal imaging problem by creating the so-called "deep neural network," which is a computer system that simulates how a human brain works. By training this network, the technology is finally able to identify a person's face by way of connecting a wide array of information available in its database, according to Tech Times.
The technology is still being developed and trained. But its accuracy - the incidences where the system identifies a match - currently stands at 80 percent, Gizmodo reported. If the image being analyzed is only one, the system can only correctly match 55 percent. The system also needs a comprehensive data in order to improve its matching capability. For the German study alone, "a set of 1,586 images from the University of Notre Dame which included pictures of 82 people with different facial expressions and in different lighting," were used according to the Daily Mail. The researchers, however, found that, this early, the smart system can outperform existing facial recognition technologies as it could identify a face in 35 milliseconds.
The new facial recognition technology is expected to be beneficial to law enforcement and covert surveillance. Dark photographs posted on Facebook, would have to rely on the site's proprietary tagging technology in the meantime.
