Is it possible to use ML and machine learning to recognize objects on satellite images?

Question

Machine learning

I saw articles on Habré where the map is automatically digitized, but for the most part it is the classification that goes there.
The question is, are there examples of exactly object recognition? Well, roughly speaking, so that the network can determine the type of aircraft on the runway. Is there such a thing? Is this even possible?
Another big question. Is it possible to say that now the neural network will decrypt in the same way or better than the decryptor?

Answer 1

Maybe. Here are the articles:
https://m.habrahabr.ru/company/ods/blog/325096/
https://m.habrahabr.ru/company/neurodatalab/blog/3...
https://m.habrahabr.ru /company/mailru/blog/335164/
https://m.habrahabr.ru/post/335026/
There was also an article from ods about vehicle recognition in pictures as part of a competition from the British, I can’t find it now.

Answer 2

Yes, it's possible. But no miracles. Enlarge the image in a drop of water, as in CSI, and you will not be able to recognize the object.
It is not always possible to determine the model of a particular aircraft, it will be rather an exception; aircraft type with high but not 100% certainty.