In this "Spot the GEO satellites" challenge, or spotGEO challenge, you are asked to develop an algorithm to detect Geostationary orbiting objects from simple png images (or frames) acquired by an unkown, low-cost ground-based telescope. Can you learn on how to cope with cloud cover, atmospheric/weather effects, light pollution, sensor noise/defects, star occlusions and more?

Check out the problem description, the dataset and the starter-kit.

This competition is organized by ESA's Advanced Concepts Team (ACT) in partnership with University of Adelaide.

News

• 11.01.2021: Full release of Version 2.0 dataset on Zenodo (containing additional annotations)
• 11.09.2020: The jury for the Kelvins out-of-the-box thinking award has decided to split the prize between the teams AgeniumSPACE and dwiuzila. Congratulations! (Details)
• 08.09.2020: The competition has concluded and the Final Ranking is available. We congratulate Team AgeniumSPACE for winning the competition!
• 01.09.2020: Submissions are closed. We thank everyone who put their hard effort in this competition. As outlined under our Rules, a final ranking will be produced for all teams that submit a .pdf-document to us until 08.09.2020. A few pages (2-3), formless are sufficient. This document should include
  • your teamname (mandatory)
  • summary of your solution approach
  • what type of software/code was used to tackle the problem
  • any form of data preprocessing or augmentation
  • if machine learning was used, which models and how they were trained
  • any remarks, observations and feedback you would like to share with the organizers
• Send your document to act "at" esa "dot" int. After the deadline, we will re-evaluate the current best submission of each team that sent us a document on the complete test-set, resulting in the final ranking and announcement of the competition winners.
• Note that we grant the "Kelvins out-of-the-box thinking" award for innovative solution approaches irrespective of your final score. Consider sending us your solution-approach as .pdf to be eligible for this award.
• Please ask any questions you might still have at the competition discussion page.
• 08.06.2020: Submissions are open! Dataset and starter kit released.
• 28.05.2020: Competition announced.