Two German scientists have created a facial recognition software that works based on infrared, so that it can identify you in the dark.
The identification of people through algorithms is a technology that is increasingly being used. Without going any further Facebook has been using it for a few years to automatically tag its users in photos. But the authorities are also using it to search for criminal suspects. Software dedicated to this task is complex, they handle a large amount of data and require great computing power. But facial recognition fails the moment the face is not fully distinguishable or there are covered parts.
Hence, specialists interested in this technology continue to investigate to save these black holes. Facebook recently announced that it could identify people even if their faces were not visible, focusing on other patterns, such as clothing. Two German scientists, who work at the Karlsruhe Institute of Technology, have gone one step further. The software they have created allows facial recognition in the dark.
In statements to the BBC, one of them pointed out that they had been investigating facial recognition for some years until they arrived at these results. System works based on infrared radiation, derived from body temperature. The software is capable of measure a person’s thermal footprint and compare it to his photograph in the light.
After analyzing a series of parameters, the system decides whether or not the captured thermal fingerprint matches the person in the photograph. The percentage of success of the software has been 80% in the tests that scientists have carried out. The researchers say that this rate could be improved if you train with more data to the program and increase the computing power used.
And although the percentage of success is high we must not forget that it is tests in a controlled environment such as the laboratory. It goes without saying the technology it is still far from having commercial applicationsbut when you go out into the real world, your success rate would be compromised. Nevertheless, it is perhaps the first time that a facial recognition system in the dark obtains these results.
Images: Micimakin and Franck Boston
