Utility function to apply conv + BN. (Slightly modified from https://github.com/fchollet/keras/blob/master/keras/applications/inception_v3.py

Helper to build a BN -> relu -> conv block. This is an improved scheme proposed in http://arxiv.org/pdf/1603.05027v2.pdf

Builds a custom ResNet like architecture. Args: input_shape: The input shape in the form (nb_channels, nb_rows, nb_cols)

This is the function used for cross channel normalization in the original Alexnet

Helper to build a BN -> relu block

Adds a shortcut between input and residual block and merges them with "sum"

Helper to build a conv -> BN -> relu block

Builds a residual block with repeating bottleneck blocks.

Creates the inception v4 network Args: num_classes: number of classes dropout_keep_prob: float, the fraction to keep before final

Given a keras [model], return an equivalent model which parallelizes the computation over [n_gpus] GPUs. Each GPU gets a slice of the input b

Basic 3 X 3 convolution blocks for use on resnets with layers <= 34. Follows improved proposed scheme in http://arxiv.org/pdf/1603.05027v2.pdf

Bottleneck architecture for > 34 layer resnet. Follows improved proposed scheme in http://arxiv.org/pdf/1603.05027v2.pdf Returns: A fi