TODO

General
=======

* embedfilters, make 3-channels images RGB/colour
* Comments/Docstrings where applicable
* Timings/Profiling
[* Initialisation methods review, pre-initialisation]
[* Add GradNet inspired stuff]
* Conv: error message?!?
[* Trainer/Config/Creation]
[* Carry FOV in tagged shapeullin m]
* Make class weights etc. shared instead of inputtable




Neural
======

[* DensePrediction]
* Mergercrop / Unet
* Inception Module
* RNN
* LSTM


Python 3 Compat
===============

Modules/Packages to Ignore
--------------------------

* data/
* TestRestore.py
* VAE.py


Other
=====

* Wrap/Improve Warping
* Graph Visualisation
* Server for Plotting
* Packaging
* Optimise Parallelisation
* Tests: transform main --> unittest