Although the unauthorized impersonation of individuals has been a constant and worrying problem of Deepic, this technology has other potentials as well. Researchers, for example, have shown that Deepick can be used to manipulate satellite imagery and create images that look real – but completely fake.
Bo Zhao of the University of Washington, who led the study, Says He does not want to scare anyone, but he wants to better show people the risks and opportunities. In fact, their approach has “light transmission” techniques rather than similar to other popularly known diphthongs.
Zhao and his team developed a machine learning system with a collection of images from three cities: Seattle, Tacoma, and Beijing. Each of these cities has its own appearance. Seattle, for example, has more vegetation and shorter streets. Beijing is more monochrome, and its tall buildings cast large shadows. The system used in this study learned to relate the details of street maps (on Google or Apple maps) to the details of satellite maps.
In the end, the system was able to show what it would look like if that area were part of the three cities by receiving each piece of street map. In the bottom two sections of the image below, you can see how the software and satellite imagery of an area is converted in combination with the geographic information of Seattle (bottom left) and Beijing (bottom right).
However, a careful examination of these simulations shows that fake maps are not as accurate as real maps. But at a glance we can give an acceptable score to this simulation. The researchers say that this method can be used to simulate images of areas where satellite imagery is not available. Also, with this mechanism, attractive images can be obtained from ancient maps, as if they were actually recorded by satellites.
The researchers note that if this technology is used in the wrong way, it is possible to carefully examine the colors and specific features of each map and determine to a good extent whether they are true or fake.