# Worldfloods: ML-based disaster detection prototype launched to space

## Team

• Gonzalo Mateo-Garcia (Universidad de Valencia)
• Joshua Veitch-Michaelis (Liverpool John Moores University)
• Lewis Smith (University of Oxford)
• Silviu Oprea (University of Edinburgh)
• Guy Schumann (University of Bristol)
• Yarin Gal (University of Oxford)
• Atılım Güneş Baydin (University of Oxford)
• Dietmar Backes (University of Luxembourg)

## Abstract

Spaceborne Earth observation is a key technology for food response, ofering valuable information to decision makers on the ground. Very large constellations of small, nano satellites— ’CubeSats’ are a promising solution to reduce revisit time in disaster areas from days to hours. However, data transmission to ground receivers is limited by constraints on power and bandwidth of CubeSats. Onboard processing ofers a solution to decrease the amount of data to transmit by reducing large sensor images to smaller data products. The ESA’s recent PhiSat-1 mission aims to facilitate the demonstration of this concept, providing the hardware capability to perform onboard processing by including a power-constrained machine learning accelerator and the software to run custom applications. This work demonstrates a food segmentation algorithm that produces food masks to be transmitted instead of the raw images, while running efciently on the accelerator aboard the PhiSat-1. Our models are trained on WorldFloods: a newly compiled dataset of 119 globally verifed fooding events from disaster response organizations, which we make available in a common format. We test the system on independent locations, demonstrating that it produces fast and accurate segmentation masks on the hardware accelerator, acting as a proof of concept for this approach.

## Publications

1. Mateo-Garcia, Gonzalo, Joshua Veitch-Michaelis, Lewis Smith, Silviu Oprea, Guy Schumann, Yarin Gal, Atılım Güneş Baydin, and Dietmar Backes. 2021. “Towards Global Flood Mapping Onboard Low Cost Satellites with Machine Learning.” Scientific Reports 11 (7249). Nature Publishing Group. doi:10.1038/s41598-021-86650-z.

Spaceborne Earth observation is a key technology for flood response, offering valuable information to decision makers on the ground. Very large constellations of small, nano satellites - ’CubeSats’ are a promising solution to reduce revisit time in disaster areas from days to hours. However, data transmission to ground receivers is limited by constraints on power and bandwidth of CubeSats. Onboard processing offers a solution to decrease the amount of data to transmit by reducing large sensor images to smaller data products. The ESA’s recent PhiSat-1 mission aims to facilitate the demonstration of this concept, providing the hardware capability to perform onboard processing by including a power-constrained machine learning accelerator and the software to run custom applications. This work demonstrates a flood segmentation algorithm that produces flood masks to be transmitted instead of the raw images, while running efficiently on the accelerator aboard the PhiSat-1. Our models are trained on \worldfloods: a newly compiled dataset of 119 globally verified flooding events from disaster response organizations, which we make available in a common format. We test the system on independent locations, demonstrating that it produces fast and accurate segmentation masks on the hardware accelerator, acting as a proof of concept for this approach.

@article{mateogarcia-2021-global,
title = {Towards Global Flood Mapping Onboard Low Cost Satellites with Machine Learning},
author = {Mateo-Garcia, Gonzalo and Veitch-Michaelis, Joshua and Smith, Lewis and Oprea, Silviu and Schumann, Guy and Gal, Yarin and Baydin, Atılım Güneş and Backes, Dietmar},
journal = {Scientific Reports},
publisher = {Nature Publishing Group},
year = {2021},
volume = {11},
number = {7249},
doi = {10.1038/s41598-021-86650-z},
url = {https://doi.org/10.1038/s41598-021-86650-z}
}

1. Mateo-Garcia, Gonzalo, Silviu Oprea, Lewis Smith, Joshua Veitch-Michaelis, Atılım Güneş Baydin, and Dietmar Backes. 2019. “Flood Detection On Low Cost Orbital Hardware.” In Artificial Intelligence for Humanitarian Assistance and Disaster Response Workshop, 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Vancouver, Canada.

Satellite imaging is a critical technology for monitoring and responding to natural disasters such as flooding. Despite the capabilities of modern satellites, there is still much to be desired from the perspective of first response organisations like UNICEF. Two main challenges are rapid access to data, and the ability to automatically identify flooded regions in images. We describe a prototypical flood segmentation system that could be deployed on a constellation of small satellites, performing processing on board to reduce downlink bandwidth by 2 orders of magnitude. We target PhiSat-1, part of the FSSCAT mission, which is planned to be launched by the European Space Agency (ESA) near the start of 2020 as a proof of concept for this new technology.

@inproceedings{mateogarcia-2019-orbital,
title = {Flood Detection On Low Cost Orbital Hardware},
author = {{Mateo-Garcia}, Gonzalo and Oprea, Silviu and Smith, Lewis and {Veitch-Michaelis}, Joshua and Baydin, Atılım Güneş and Backes, Dietmar},
booktitle = {Artificial Intelligence for Humanitarian Assistance and Disaster Response Workshop, 33rd Conference on Neural
Information Processing Systems (NeurIPS 2019), Vancouver, Canada},
year = {2019}
}


## Acknowledgments

Tis research was conducted at the Frontier Development Lab (FDL), Europe, 2019 led by Trillium Technologies. Te authors gratefully acknowledge support from European Space Agency (ESA); Pierre-Philippe Mathieu and the team at PhiLab at ESA ESRIN; Google LLC; Intel Corporation; Kellogg College, University of Oxford; UNICEF representative Dohyung Kim.; James Parr, Jodie Hughes, Leo Silverberg and Sarah McGeehan from Trillium Technologies and other organisations and mentors who supported FDL Europe 2019. G.M.-G. has been supported by the Spanish Ministry of Science and Innovation (Projects TEC2016-77741-R and PID2019-109026RBI00, ERDF). S.O. is supported by the EPSRC Grant EP/L016427/1 and Te Financial Times. A.G.B. is supported by EPSRC/MURI Grant EP/N019474/1 and by Lawrence Berkeley National Lab. J.V.M. is supported by STFC/GCRF Grant ST/R002673/1.