Soteria

We won Best Solution to AWS Disaster Response Hackathon! 🥳

Featured in Amazon re:MARS 2022 - Improving disaster response with machine learning

Soteria uses machine learning with satellite imagery to map natural disaster impacts for faster emergency response.

Youtube Demo: https://youtu.be/frjIm_FDlhc
Devpost Home page: https://devpost.com/software/soteria-yolciw
Hugging Face ML Demo: https://huggingface.co/spaces/samt/soteria
Figma Prototype: link
Presentation Slides: link

ML Models

Binary Damage Classification:
Disaster Type Classification:
Regional Damage Level Classification:

Our Team

Background

The scale, scope and intensity of natural disasters ranging from hurricanes to wildfire is only increasing as the effects of climate change worsen. The lives lost and impacted continue to highlight peoples vulnerability to these disastrous events. As a team, we wanted to use our areas of interest and expertise to serve communities who have or will be impacted by natural disasters. We don’t need to be on the ground of a disaster to make an impact. Inspired by the potential that AI has for improving the quality of life, we applied this to natural disasters. We wanted our model to be applicable to all natural disaster globally, but first we start on the East coast of Malaysia.

Our Project

Technologies

Machine Learning

Dataset

Download: https://xview2.org/ Learn more: https://arxiv.org/abs/1911.09296

Models

Figma Prototype

Inspirations/Resources

• AI For Good — Disaster Response by Shairoz Sohail
• PEER HUB IMAGENET Φ-NET
• Building Damage Detection in Satellite Imagery Using Convolutional Neural Networks by Google
• Structural Damage Image Classification