rhythm92 / wide_resnets_keras

Keras (using Theano) implementation of "Wide Residual Networks"

Keras (Theano) implementation of "Wide Residual Networks"

This repo contains the code to run Wide Residual Networks using Keras running Theano.

• Paper: http://arxiv.org/abs/1605.07146
• Original code: https://github.com/szagoruyko/wide-residual-networks

Dependencies:

• Keras, numpy - can be installed with sudo pip install -r requirements.txt.
• Theano; install latest version with sudo pip install git+git://github.com/Theano/Theano.git --upgrade --no-deps.
• To plot the architecture of the model used (like the plot of the WRN-16-2 architecture plotted below), you need to install pydot and graphviz. I had to install the C-version for graphviz first (following comments in this issue):

$ sudo apt-get install graphviz
$ sudo pip install pydot
$ sudo pip install graphviz

Usage Details:

Run the default configuration (i.e. best configuration for CIFAR10 from original paper/code, WRN-28-10 without dropout) with:

$ python main.py

There are three configuration sections at the top of main.py:

• NETWORK CONFIGURATION: Includes the main parameters the authors experimented with.
• DATA CONFIGURATION: Containing data details.
• OUTPUT CONFIGURATION: Defines paths regarding where to save model/checkpoint weights and plots.

Results:

• Running WRN-40-4 no dropout now.

##WRN-16-2 Architecture