bhrnjica/anndotnet: ANNdotNET v1.2 - deep learning tool on .NET platform

This version brings several important updates and new features:

The following enhancements and new features has been made in this release

• ### Image classification module
  • Image Classification project - This is huge step forward to this version. With IC project, the user can completely implements image classification with minimal steps without leaving GUI module.#39
  • ANNdotNET Feed - brings ability to share interesting annproject's to all community. Once the interesting project is added to ANNdotNET Feed it can be viewed by all uses that have installed ANNdotNET v1.2+ version.#33
• ### Raw Data Handling Module
  • Time Series Generator - previously time series could be loaded in Data Import dialog only with one column data without header. Now, more than one column with header can be imported, and only the last one will be generated as time series, while the rest columns will remain as are.#34 #35
  • Split Raw Data Set to: train, validation and test sets. Up to now the user could split raw data set on train and validation sets only.#22
  • Export to Excel with all three data sets.#23
  • Optimization data loading and handling with huge data set.#21

• ### Neural Network Designer improvements

  • Add new Layer types: Convolution, Pooling, etc. #18
  • Insert button Insert Layer in the network at specific position.#36

Upgrade to CNTK 2.7

• All modules are upgraded to CNTK 2.7.
• CUDA 10 must be installed in order to support GPU processing.

Bug Fixes

• Some WinForms Dialogs has been converted into WPF based windows in order to fix WinForm DPI issue on scaled monitors. #42
• Minor code improvements and bug fixes