Announcing the Launch of the AI Model Challenge for Disaster Response
We are thrilled to announce the launch of the Charter AI Challenge for Earthquake Response, a groundbreaking initiative of the International Charter ‘Space & Major Disasters’ in collaboration with the European Space Agency’s ESA Phi-Lab and other partners from the EO community such as LIST, Sinergise, Earthpulse, NovaSpace, and Gmatics.
This global challenge is a call to action for data scientists, AI researchers, and disaster response specialists to develop cutting-edge artificial intelligence models capable of rapidly detecting and mapping earthquake damage at the building level using high-resolution satellite imagery. The data to be provided to the participants is archived satellite imagery from past earthquake activations of the Charter dating back to 2020.
In the aftermath of major earthquakes, timely and accurate damage assessments are critical for saving lives and coordinating relief efforts. Satellite Earth Observation (EO) offers a powerful vantage point, but current tools fall short of reliably mapping damage at the building level. This challenge seeks to change that by providing participants with a large set of multi-mission, high-resolution data to train their AI model.
By harnessing the power of AI, participants will help revolutionize how EO data is used during disasters. This aims to bring greater speed, accuracy, and scalability to humanitarian response worldwide.
What’s Involved?
The challenge is open to participants worldwide, from students and startups to research institutions and industry leaders. ESA and CNES data scientists will offer guidance and technical assistance throughout the challenge.
Participants will:
- Develop AI models that identify damaged buildings from satellite imagery
- Use data from past Charter earthquake activations
- Compete for a cash prize and the opportunity to attend the next Charter meeting to present their model to the board
- The winning model(s) will be considered for integration as a processing service in the ESA Charter Mapper environment
The focus of this year’s thematic track is Earthquake Response, with EO data sourced from key global events including Türkiye, Syria, Morocco, Afghanistan, and more. The dataset includes a selection of pre- and post-event satellite imagery acquired from BlackSky Global, Gaofen-2, GeoEye-1, Kompsat-3, Kompsat-3A, Pléiades-1, WorldView-2, and WorldView-3 satellites. This dataset provides high resolution calibrated optical EO data provided by multiple space agencies during Charter activations and processed by the ESA Charter Mapper platform.
The Challenge will have two phases: Phase 1 is an 8-week training phase with integrated live scoring, and Phase 2 is the 10-day stress test that will take place at the end of the challenge, with unseen testing data. Phase one will be 40% of the overall score with Phase 2 taking 60%.
What’s at Stake?
Winning models will be promoted in open-science forums, with the potential for integration into the ESA Charter Mapper, the cloud-based EO processing environment of the Charter. The integration of a successful model into the Charter ecosystem would offer international disaster management agencies a powerful new tool for real-time decision making during emergencies.
This is more than a competition. It’s a collaborative effort to advance science, technology, and humanitarian impact through space-based innovation. Together, let’s transform disaster response, one algorithm at a time.
Challenge Start: 11 July 2025
Challenge End: 29 September 2025
Announcement of winner(s): At the 54th Charter meeting in Strasbourg 6th – 10th October 2025
Registration and additional info: Φ-Lab Challenges