Signature Verification

Setup Environment

In order to prevent package conflicts for different projects, it is recommended to install a python environment manager such as virualenv. Because of its rich features I personally use anaconda which provides similar functionallity to virtualenv with more features such as creating environment with different python versions without actually installing it globally on system. For more information on installing anaconda see the documentations. (Since anaconda has many preinstalled packages which might not be needed, you could install the minimal version, miniconda)

If you prefer to use virtualenv you could find the installation instruction in documentations.

conda create -n sigv python=3  # Create a python 3 environment with the name of sigv

Install requirements

source activate sigv  # Activate project environment
conda install ipython jupyter matplotlib numpy h5py  # Install packages available on conda repository
pip install keras  # Insallt packages available on pypi repository

Install tensorflow from source

It is possible to install both CPU and GPU versions of tensorflow via pip but because of the optimizations and customizations which are available, it is recommended to install tensorflow from source. For more information on configuring tensorflow installation see the documentations.

git submodule update --init
cd libs/tensorflow
./configure
bazel build --config=opt //tensorflow/tools/pip_package:build_pip_package
bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
pip install /tmp/tensorflow_pkg/[tensorflow_package] 

Configuration

The configuraion must be in configuration.yaml. For a sample configuration see the provided sample here.

Sample configuration

If you want to use the pre-written configuration, you could copy the provided sample configuration.sample.yaml.

rsync -a --ignore-existing configuration.sample.yaml configuration.yaml