Crop Map Generation

End-to-end workflow for generating high resolution cropland maps.

Contents

• Creating a crop map
• Training a new model
• Setting up a local environment
• Adding new labeled data
• Tests
• Previously generated crop maps
• Acknowledgments
• Reference

Creating a crop map

To create a crop map run the following colab notebook (or use it as a guide):

Training a new model

To train a new model run the following colab notebook (or use it as a guide):

Two models are trained - a multi-headed pixel wise classifier to classify pixels as containing crop or not, and a multi-spectral satellite image forecaster which forecasts a 12 month timeseries given a partial input:

Adding new labeled data

To add new labeled data run the following colab notebook (or use it as a guide):

Setting up a local environment

Ensure you have anaconda installed.

1. For development

Ensure you have gcloud installed.

conda install mamba -n base -c conda-forge  # Install mamba
mamba env create -f environment-dev.yml     # Create environment with mamba (faster)
conda activate landcover-mapping            # Activate environment
gcloud auth application-default login       # Authenticates with Google Cloud

2. For shapefile notebook

conda env create -f environment-lite.yml    # Create environment
conda activate landcover-lite               # Activate environment
jupyter notebook

Tests

The following tests can be run against the pipeline:

flake8 . # code formatting
mypy .  # type checking
python -m unittest # unit tests

# Integration tests
python -m unittest test/integration_test_labeled.py
python -m unittest test/integration_test_model_bbox.py
python -m unittest test/integration_test_model_evaluation.py

Previously generated crop maps

Google Earth Engine:

• Kenya (post season)
• Busia (in season)

Zenodo

• Kenya (post season) and Busia (in season).

Reference

If you find this code useful, please cite the following paper:

Gabriel Tseng, Hannah Kerner, Catherine Nakalembe and Inbal Becker-Reshef. 2020. Annual and in-season mapping of cropland at field scale with sparse labels. Tackling Climate Change with Machine Learning workshop at NeurIPS ’20: December 11th, 2020