Researchers from the Catholic University of Leuven in Belgium have published a scientific report on the deception of video surveillance systems. As it turns out, a simple 2D image printed on a T-shirt or bag can make a person invisible to security cameras if they rely on machine learning for their work and must recognize people in the video stream.
To achieve the desired effect, an image of 40x40 centimeters (which is designated by the word patch in the experts' report) should be located in the middle of the camera's detection box and be constantly in its field of view. Of course, this method will not help a person to hide a face, however, the algorithm for detecting people, in principle, will not be able to detect a person in the frame, which means that the subsequent recognition of facial features will also not be launched.
In the course of experiments, the researchers tried to use a variety of images to deceive surveillance systems, including abstract "noise" and blurred images, but it turned out that photographs of random objects that underwent various processing were best suited. For example, in the illustration below, you can see "patches" created from random images that were rotated 20 degrees, scaled randomly, added "noise", and also randomly modified brightness and contrast.
If you put the resulting pictures on clothes, a bag, and so on, the algorithms will no longer see the person behind them. You can clearly see the effectiveness of such pictures by watching the PoC video published by experts. The experts tested their method on the open-source Darknet neural network, which uses the YOLOv2 (You Only Look Once) real-time object detection system.
In the same way, it is possible to "hide" not only a person from the cameras, but also any other object, for example, the surveillance system "will not see" a car or a bag with a "patch" applied to it ...
In addition to the publication of the report and the video, the researchers posted on GitHub the source codes that they used to create patches, so anyone can repeat and continue their experiments.
