Work in progress Currently this function seems to break backpropagation in TF, it seemed to have worked once or twice but I can't get it to work properly, backpropagation may need to be formulated explicitly ( this is currently out of my expertise)
As of now the package needs tensorflow and numpy to work, matplotlib is useful for the plotting scrip.
File explanations:
chainerutils: code from chainer with some modified functions to split the data reshaper: testing ground for the TF implementation ( and potentially MXNET and CNTK) varitest: implementation of the baseline functions and potential functions for chainer varplot: currently use to plot the two functions of varitest in order to see how the bias affects the output edger: makes kernels for testing proof_of_concept: make plot of images to compare the convolution from imgaug, this method and the convs from scipy and tensorflow
#Proof of concept demonstrations are in the POC folder