Desk-AId: Humanitarian Aid Desk Assessment with Geospatial AI for Predicting Risk of Landmines

This repository contains the code, data, and supplementary materials for the research paper:

"Desk-AId: Humanitarian Aid Desk Assessment with Geospatial AI for Predicting Risk of Landmines"
Authors: Flavio Cirillo, Guerkan Solmaz, Yi-Hsuan Peng, Christian Bizer and Martin Jebens

F. Cirillo, and G. Solmaz are with NEC Laboratories Europe, Heidelberg, Germany (e-mail: {flavio.cirillo;gurkan.solmaz}@neclab.eu) Yi-Hsuan Peng and Christian Bizer are with University of Mannheim, Mannheim, Germany Martin Jebens is with Copenhagen Hazard Mapping, Copenhagen, Denmark,

Pre-print:
https://doi.org/10.48550/arXiv.2405.09444

---

Abstract

The process of clearing areas, namely demining, starts by assessing and prioritizing potential hazardous areas (i.e., desk assessment) to go under thorough investigation of experts, who confirm the risk and proceed with the mines clearance operations. This paper presents Desk-AId that supports the desk assessment phase by estimating landmine risks using geospatial data and socioeconomic information. Desk-AId uses a Geospatial AI approach specialized to landmines. The approach includes mixed data sampling strategies and context-enrichment by historical conflicts and key multi-domain facilities (e.g., buildings, roads, health sites). The proposed system addresses the issue of having only ground-truth for confirmed hazardous areas by implementing a new hard-negative data sampling strategy, where negative points are sampled in the vicinity of hazardous areas. Experiments validate Desk-Aid in two domains for landmine risk assessment: 1) country-wide, and 2) uncharted study areas). The proposed approach increases the estimation accuracies up to 92%, for different classification models such as RandomForest (RF), Feedforward Neural Networks (FNN), and Graph Neural Networks (GNN).

---

Repository Structure

├── dataset_input/ # Raw and pre-processed dataset used in the code
├── result/ # Outcome of the jupyter and python scripts used in the paper
├── *.ipynb # Jupyter notebooks for analysis and visualization
├── FNN_GNN.py # Python script to train and test neural networks
├── README.md # Project documentation (this file)
└── LICENSE.txt # License for use and distribution

What to Run

• 1.Random Sampling vs Hard Negative Sampling_usingReadAndPrepare.ipynb

  This Jupyter notebook generates the results contained in section 6.1 of the pre-print Desk-AId paper

• 2.HardNegative_18 features_usingReadAndPrepare.ipynb

  This Jupyter notebook generates the results contained in section 6.3 of the pre-print Desk-AId paper

• 3.Surrounding vs Study Area.ipynb

  This Jupyter notebook generates the results contained in section 6.2 of the pre-print Desk-AId paper

• 4.FNN_GNN.py

  This Python script generated the results contained in section 6.4 of the pre-print Desk-AId paper

Other Files:

• read_and_prepare_datasets.ipynb

  This file is an utility file used by the jupyter notebook 1-3

How to Cite

@article{cirillo2024desk,
  title={Desk-AId: Humanitarian Aid Desk Assessment with Geospatial AI for Predicting Landmine Areas},
  author={Cirillo, Flavio and Solmaz, G{\"u}rkan and Peng, Yi-Hsuan and Bizer, Christian and Jebens, Martin},
  journal={arXiv preprint arXiv:2405.09444},
  year={2024}
}