geospatial-learn
Geospatial-learn is a Python lib for using scikit-learn, xgb and keras models with geo-spatial data. Some raster and vector manipulation is also included. The aim is to produce convenient, relatively minimal commands for putting together geo-spatial processing chains and using machine learning (ML) libs. The name is a play-on of scikit-learn, though I really ought to think of something better. The functions are mainly a collection resulting from my own research area of remote sensing and image processing; hence some of it may be esoteric but there are some fairly typical processing tasks too.
- There is a 'raster' module for (you guessed it) raster/image processing. This is not exaustive set of functions of course, just things that have been convenient and repeated such as I/O, masking, some filtering.
- There is a 'shape' module for vector processing which is mainly based around extracting image properties and writing them to a vector format. Functions include zonal stats, glcm-based texture etc as well as further manipulation of lines and polygons using things like active contours.
- The 'learning' module is for applying creating ML models and applying them to raster, vector and point cloud data. This is all based around sklearn, xgboost, keras and t-pot.
- The 'handyplots' module contains a few simple functions that may be useful e.g. plot a classifcation report, confusion matrix etc.
- The 'utils' module is full of stuff which is yet to be given a home that makes sense and may not all be doc'd - take a look...
Dependencies
geospatial-learn requires:
- Python 3
- Anaconda
User installation
- Installation uses the anaconda/miniconda system - please install this first if you don't have it already
- Clone the repository or download and unzip the tar
- cd into the folder and type the following
conda env create -f geolearn_env.yml
Alternatively, for a shorter wait (conda is quite slow these days), the mamba system is recommended, install this in your base conda then:
mamba env create -f geolearn_env.yml
- To activate - type
conda activate geospatial_learn
Quickstart
A summary of some functions can be found here:
https://github.com/Ciaran1981/geospatial-learn/blob/master/docs/quickstart.rst
There are some jupyter-based workflows to be found here:
https://github.com/Ciaran1981/geospatial-learn/tree/master/example_notebooks
These are currently a work in progress of course!
Docs
Documentation can be found here:
https://ciaran1981.github.io/geospatial-learn/docs/html/index.html
These are a work in progress!
Development
Any contributors of all experience levels are welcome
Project History
The functionality was written as part of various research projects involving Earth observation & geo-spatial data.
Citation
If you use geospatial-learn in a scientific publication, citations would be appreciated - click on the blue zenodo link at the top.
Robb, C. (2017). Ciaran1981/geospatial-learn: Geospatial-learn 0.3 release. Zenodo. https://doi.org/10.5281/ZENODO.3968431
A .bib file is available in the repo (geolearn.bib)