GitLab

* Use the tensorflow cross entropy function to prevent nan losses.

* Correct double softmax and use mean for loss.

When importing tensorflow before scipi.ndimage and running the
example in a docker container, the following error is given:

```
 % python example.py
Traceback (most recent call last):
  File "example.py", line 25, in <module>
    im = im / 255.
TypeError: unsupported operand type(s) for /: 'instance' and 'float'
```
This only happens for 'cat.jpg'. When converting the image to a '.png'
file, the example runs as expected.

When swapping the imports around so that scipy.ndimage is imported
BEFORE tensorflow, both 'jpg' and 'png' files work.

* Changed the examples to specify the output size instead of the downsample_factor.

This is required by PR #57

* Address flake8 errors.

* Update readme and parameter descriptions.

* Modified the way the output size is specified.
* Added support for batches of inputs.

Shorten STN summary in README

relinked to data files

adding license header, editing AUTHORS file

adding tensorflow version