This code was developed by JHU/APL to demonstrate how artificial intelligence and machine learning can be leveraged IDP and Refugee Camp planning and safety monitoring

If you use our code, please cite our paper:

@article{hadzic2020estimating,
  title={Estimating Displaced Populations from Overhead},
  author={Hadzic, Armin and Christie, Gordon and Freeman, Jeffrey and Dismer, Amber and Bullard, Stevan and Greiner, Ashley and Jacobs, Nathan and Mukherjee, Ryan},
  journal={arXiv preprint arXiv:2006.14547},
  year={2020}
}

For details and instructions for downloading the dataset used in this work, please see dataset.

See requirements.txt.

Start by modifying the path information in params.py to ensure that it is valid, pointing to the location where you have downloaded and formatted the IOM data.

Run the code using:

python main.py

• --train to train the model
• --test to evaluate the model performance
• --augmentation to perform data augmentation during training
• --multiprocessing leverage multiple processes to speed up dataloading
• --add-osm use OSM structure mask data
• --gpu [GPU_ID(s)] specify GPU device IDs to use