Automated classification of land cover using CNNs

Table of contents:

• All code is in modules in cnn-land-cover/scripts/
• Example notebooks of how to use these modules is in cnn-land-cover/notebooks/

Installation:

• geo.yml contains all python package versions used for this repo. Create a new virtual environment with all these packages with Anaconda by typing in terminal: conda env create -f geo.yml and then conda activate geo.
• Next, install final packages using pip: pip install geocube==0.1.0, pip install patchify, pip install segmentation-models-pytorch, pip install torchsummary, pip install pytorch==1.12.1+cu102 -f https://download.pytorch.org/whl/torch_stable.html, pip install torch-tb-profiler, pip install -U kaleido
• User-specific file paths are stored in content/data_paths.json.

Interpretation key:

• Please see https://reports.peakdistrict.gov.uk/interpretation-key/docs/introduction.html