Need to find ways of using theno backend for Keras rather than tensorflow

[compressionmonkey]

Could loading theano backend be better than running tensorflow to make visualization?