Vison307 / DeepST

Deep Learning for Spatio-Temporal Data

DeepST

DeepST: A Deep Learning Toolbox for Spatio-Temporal Data

Tested on Windows Server 2012 R2.

Installation

DeepST uses the following dependencies:

• Keras and its dependencies are required to use DeepST. Please read Keras Configuration for the configuration setting.
• Theano or TensorFlow, but Theano is recommended.
• numpy and scipy
• HDF5 and h5py
• pandas
• CUDA 7.5 or latest version. And cuDNN is highly recommended.

To install DeepST, cd to the DeepST folder and run the install command:

python setup.py install

To install the development version:

python setup.py develop

Data path

The default DATAPATH variable is DATAPATH=[path_to_DeepST]/data. You may set your DATAPATH variable using

# Windows
set DATAPATH=[path_to_your_data]

# Linux
export DATAPATH=[path_to_your_data]

License

DeepST is released under the MIT License (refer to the LICENSE file for details).