Exam project in the course "Geospatial Data Science" Spring 2023. Project was developed by Aske Schytt Meineche and Viktor Due Pedersen.
Training this model is a process that takes a long time. Therefore, we have included the trained model in the repository under model_outputs. For the same reason, we haven't included any data but instead provided instructions on how to download the data used in this project in biomassters-download-instructions.txt. In the folder large_samplewe have included a sample of the data used in this project. Note that training can be done by running the train_CNN.py script. This will train the model and save it in the model_outputs folder, but it will take a long time, even on the HPC. If you decide to train the model yourself, there are some packages that needs to be installed that are not specified in the requirements.txt file. The required packages are in the import statements in the files.
This is the file that takes care of all the evaluation and presentation of the results. It is a jupyter notebook that can be run from top to bottom. It will then produce all the figures and tables used in the report.
Some of the files in this repository are copies of files from other folders in this repository. This is due to the fact that we wanted the training folder to be self-contained.