geospatial-learn ============

geospatial-learn is a Python module for using scikit-learn and xgb models with geo-spatial data, chiefly raster and vector formats.

The module also contains various fuctionality for manipulating raster and vector data as well as some utilities aimed at processing Sentinel 2 data.

The aim is to produce convenient, minimal commands for putting together geo-spatial processing chains using machine learning libs. Development will aim to expand the variety of libs/algorithms available for machine learning beyond the current complement.

geospatial-learn requires:

• Python 3

At present the setup.py only installs some of the dependencies. An anaconda package is in the works, but until that is done please do the following. This assumes you have an anaconda installation with a python 3 root OR env.

Library & pypi install

download from here:

https://github.com/Ciaran1981/geospatial-learn/raw/master/dist/geospatial-learn-0.131.tar.gz

• cd into the folder
• open a terminal and type:

python setup.py install

This will install the library and packages unavailable on anaconda.

Step 2.

Conda is very handy at managing packages, hence the 2 stage install, as some of these are external to python or themselves have multiple depends.

Next, type the following (in the same terminal).

chmod +x install_conda_packages.sh

bash ./install_conda_packages.sh

All the appropriate anaconda packages will then install

Commiserations, you are using Windows (hehe). This seems to work, though I have only tested on 1 machine.

Library & pypi install

Step 1.

download from here:

https://github.com/Ciaran1981/geospatial-learn/raw/master/dist/geospatial-learn-0.130.tar.gz

• cd into the folder
• open a powershell and type:

python setup.py install

This will install the library and packages unavailable on anaconda.

Step 2.

Conda is very handy at managing packages, hence the 2 stage install, as some of these are external to python or themselves have multiple depends.

Next, type the following (in the same terminal).

.\install_conda_packages.bat

If you run into problems here, such as certain packages unavailable with Python 3.5/6, I suggest creating a conda environment with python 3.4, then following the above procedure. At the time of writing for example (31/08/17), gdal is not available in py3.5+ on windows anaconda and py3.6 on linux platforms.

I have not provided xgboost instructions here, there are some on the native website along with ensuring the lib points to your python environment of choice.

A summary of some functions can be found here:

https://github.com/Ciaran1981/geospatial-learn/blob/master/docs/quickstart.rst

This is currently a work in progress of course!

Documentation can be found here:

http://geospatial-learn.readthedocs.io/en/latest/

These are a work in progress!

New contributors of all experience levels are welcome

Here are some links to the principal libs used in geospatial-learn.

https://github.com/scikit-learn/

http://xgboost.readthedocs.io/en/latest/

http://scikit-learn.org/stable/

http://www.gdal.org/

http://www.numpy.org/

https://www.scipy.org/

http://scikit-image.org/

available soon

Geospatial-learn was originally written by Dr Ciaran Robb, University of Leicester. The functionality was written as part of various research projects involving Earth observation & geo-spatial data.

Geospatial-learn is currently written and maintained by Ciaran Robb and John Roberts. The module is at a very early stage at present and there is more material wrtten that has yet to be added.

available soon

If you use geospatial-learn in a scientific publication, citations would be appreciated