Cool NN and DL applications

Deep learning and Neural Networks have some really cool applications (ideas), and here are a few I have come across and are really tangential to mainstream applications.

Defeating Image Obfuscation with Deep Learning

This paper from Cornell and University of Texas at Austin uses DL to reconstruct the images which are pixelated, blurred or encrypted using privacy-preserving photo (P3) algorithm. These techniques are many times used to hide the identities in some sensitive videos and photos in various media. These techniques are as well used in cryptography and hence current DL methods will push cryptographes to discover methods to ensure privacy of media can be maintained.

Detecting the Programming Language of Source Code Snippets using Machine Learning and Neural Networks

I came across this blog post while searching for something related to compilation. I found this idea very appreciative and that can be implemented by multiple code engines like codechef, hackerrank, etc. It is a well thought and built approach and does not even use deep neural nets for the purpose. They have also hosted a trial platform where you can test their model.

Deep Gold

Deep learning on images has been around for sometime right now and we have seen some really impressive results. Similarly for finding locations that may have minerals or oil, satellite images are used, they are judged by geologists, who give a conclusion. The author here has developed a convolutional net that when trained, tells you the possible locations of a goldmine field. Though the results are not really impressive, but it has shown what more we can expect from DL in next few years. DL to find Oil is coming soon.