ChaBuD2023

A Machine Learning data pipeline for the Change detection for Burned area Delineation (ChaBuD) challenge at the ECML/PKDD 2023 conference.

Getting started

Quickstart

Launch into a JupyterLab environment on

Binder	SageMaker Studio Lab	Planetary Computer

Installation

Basic

To help out with development, start by cloning this repo-url

git clone <repo-url>

Then I recommend using mamba to install the dependencies. A virtual environment will also be created with Python and JupyterLab installed.

cd chabud2023
mamba env create --file environment.yml

Activate the virtual environment first.

mamba activate chabud

Finally, double-check that the libraries have been installed.

mamba list

Advanced

This is for those who want full reproducibility of the virtual environment. Create a virtual environment with just Python and conda-lock installed first.

mamba create --name chabud python=3.11 conda-lock=2.0.0
mamba activate chabud

Generate a unified conda-lock.yml file based on the dependency specification in environment.yml. Use only when creating a new conda-lock.yml file or refreshing an existing one.

conda-lock lock --mamba --file environment.yml

Installing/Updating a virtual environment from a lockile. Use this to sync your dependencies to the exact versions in the conda-lock.yml file.

conda-lock install --mamba --name chabud conda-lock.yml

See also https://conda.github.io/conda-lock/output/#unified-lockfile for more usage details.

Usage

Running jupyter lab

mamba activate chabud
python -m ipykernel install --user --name chabud  # to install virtual env properly
jupyter kernelspec list --json                    # see if kernel is installed
jupyter lab &

Running the model

The neural network model can be ran via LightningCLI v2. To check out the different options available, and look at the hyperparameter configurations, run:

python trainer.py --help
python trainer.py test --print_config

To quickly test the model on one batch in the validation set:

python trainer.py validate --trainer.fast_dev_run=True

To train the model for a hundred epochs and log metrics to WandB:

python trainer.py fit --trainer.max_epochs=100 \
                      --trainer.logger=WandbLogger \
                      --trainer.logger.entity=devseed \
                      --trainer.logger.project=chabud2023

To generate the CSV file of predicted masks on the validation set for submission to https://huggingface.co/spaces/competitions/ChaBuD-ECML-PKDD2023.

python trainer.py test --model.submission_filepath=submission.csv

More options can be found using python trainer.py fit --help, or at the LightningCLI docs.