The neural network draws cataclysms anywhere in the world. You can try your home

Canadian scientists from the Quebec Institute of Artificial Intelligence launched the site ThisClimateDoesNotExist to illustrate public awareness of the problems of natural disasters. The essence of the service is quite simple – the user needs to select a point on the world map, the site will load a photo from the specified location, and then the neural networks will redraw the picture as if the photo captured a flood, forest fire or smog.

This is what the web application interface looks like

The service uses Google Street View to search for images, so the accuracy of the maps is extremely high. However, the catch is that you cannot rotate the camera in the built-in viewer, which complicates the selection of the desired frame.

The processed images are available in 640 × 640 pixels – most often the result is mediocre, especially if people are captured on it. But sometimes GAN-neural networks successfully cope with their task.

All stages of a neural network when drawing a flood
All stages of the neural network when painting a fire
All stages of the neural network when painting smog

You may also like