Delineating agricultural field boundaries from Sentinel-2 imagery. Code from Jesse Bakker's MA thesis project, University of Minnesota Department of Geography, Environment and Society.
Functions and descriptions are in the fields_functions.py and parameters are saved in the global_config.py file. The Fields Code Processing Demo notebook walks through the processing steps. You will need to point the 'prep_file_dir' parameter in the global_config dictionary to a local directory with zipped .SAFE files from Sentinel-2 or use the sample data.
To create the conda environment to run the code, use the environment.yml file:
conda env create --file environment.yml
This should work on Mac or Linux OS, but for Windows you may encounter an error trying to create the environment this way. Instead, you can manually create the environment with the following commands:
- Create an empty environment without any packages:
conda create -n geoenv - Activate the new environment:
conda activate geoenv - Install all the necessary packages at once so that conda handles the dependencies:
conda install -c conda-forge python=3.8.3 gdal=3.0.4 geopandas=0.7.0 rasterio=1.1.5 dask=2.19.0 xarray=0.15.1 matplotlib=3.2.1 seaborn=0.10.1 scikit-learn=0.23.1 scikit-image=0.17.2 ipython=7.15.0 ipykernel=5.3.0 folium=0.11.0 bokeh=2.1.1 holoviews=1.13.3 datashader=0.11.0 psycopg2=2.8.5 sqlalchemy=1.3.17 geoalchemy2=0.6.3 descartes=1.1.0 contextily=1.0.0 memory_profiler=0.57.0 autopep8=1.5.3 netcdf4=1.5.3