The required packages and libraries are all in terence_env.yml
My own enviroment:
OS: GNU/Linux 3.16.0-4-amd64 x86_64 Debian.jessie
python version: anaconda 1.4.0
python 2.7.12

The Street View House Numbers (SVHN) Dataset

Make sure you have installed anaconda

1. Create a local enviroment as mine:
   conda env create -f terence_env.yml

2. Activate the enviroment(whose name is 'tensorflow'):
   source activate tensorflow

jupyter notebook data_preprocess.ipynb

.
├── conv_net.ipynb
├── convonet_model.mdl
├── convonet_model.mdl.meta
├── data_preprocess.ipynb
├── extra/
├── extra.tar.gz
├── project_report.pdf
├── README.md
├── summaries
├── SVHN.pickle
├── terence_env.yml
├── test/
├── test.pickle
├── test.tar.gz
├── toy_data.pickle
├── train/
├── train.tar.gz
└── valid.pickle