Hi Alexey,

deep neural network works well in most areas. I suggest that you read articles about the use of convolutional neural network and the Autoencoder and Perceptron Networks for remote sensing.

I using Autoencoder and convolutional neural networks in the field of remote sensing. If you have good hardware, you can get good results.

Any type of neural network works well for a topic.

You need to specify the theme of the caritone then look for the best neural network in the subject.

Good luck

Hi Kaveh,

Thanks for the recommendation. I'd follow your suggestion. The point is that I'm working on one open-source toolkit (http://igis.tk). Since it uses the Tcl/Tk programming language, I'm looking for a deep learning framework to call it from within the Tcl code. I briefly examined all major deep learning frameworks and concluded that I could probably write some bindings for MXNet. It seems much more difficult with others. In the beginning, I wanted to use FANN. But, it could be not such interesting for our potential users and myself. Thus, I'm asking about add-hock achievements in Remote Sensing with MXNet. Thanks