DeepFore[st]cast'

Leveraging deep convolutional neural networks to forecast tropical deforestation.

Citation

Please cite:

Thorstenson, R. (2024). Forecasting Deforestation in India with Deep Learning for the GREEN Meghalaya PES Program. Yale University.

Ball, J. G. C., Petrova, K., Coomes, D. A., & Flaxman, S. (2022). Using deep convolutional neural networks to forecast spatial patterns of Amazonian deforestation. Methods in Ecology and Evolution, 13, 2622– 2634. https://doi.org/10.1111/2041-210X.13953

Requirements

• Python 3.8+
• scikit-learn
• torch 1.9.0
• torchaudio 0.9.0
• torchvision 0.10.0

See src/requirements/environment.yml for the complete list.

Getting started

First, create the directories and download the appropriate data. See https://github.com/PatBall1/DeepForestcast/tree/master for details. Also see the Makefile in the root, src/bash_scripts/run_data_load.sh, and src/main_data_load.py.

Then, update the .env file. See src/.env.example. Depending on your use-case, not all are required.

Next, build a config in src/configs.py, then call python3 src/models.py. It can take the parameters below. If preferred, use the bash scripts src/bash_scripts/.

--debug: Enables debugging mode by overriding model settings for a simpler and faster configuration. Disables Weights & Biases integration if --wandb_project is specified.

--skip_wandb: Disables Weights & Biases logging for the run.

--test_only: Skips training and runs only testing. Requires a pretrained model specified via --init_model.

--wandb_project: Specifies a custom W&B project for logging.

--init_model (str): Path to a pretrained model to initialize training or testing. The file should have a .pt extension.

--test_epochs (int): Specifies how often (in terms of epochs) to test the model during training.

--config (str): Path to the configuration file for the model.

--config_object (str): Specifies a Python object within the file provided by --config to load the model configuration.