• About
  • Competitions
  • The OPS-SAT case
  • Register
  • Sign In
    Forgot password?
    GitHub
    By logging in with Github you accept our Privacy Policy
The OPS-SAT case
  • Home
  • Challenge
  • Data
  • Scoring
  • Rules
  • Submission Rules
  • Leaderboard
  • Results
  • Discussion
  • Resources
Timeline

July 1, 2022, midnight UTC

Oct. 31, 2022, midnight UTC

The competition is over.

→ AI on the edge: "the OPS-SAT case"

One satellite, eight classes ... and only a few shots!

In space you must live with what you have! In the data-centric machine learning challenge "the OPS-SAT case" you are given a (quantized) neural model that has passed all the requirements for inference on-board the European Space Agency's satellite OPS-SAT. Can you train its parameters as to predict one of eight classes for some tile coming directly from the spacecraft imaging sensor? .... ah, there is a catch, you will not have as many labels to work with as you would like!

On the bright side, should you be brave enough and succeed in this few-shots learning, quantization aware challenge, you will be given access to the satellite, your model uploaded and used for inferences during a real time, dedicated observational campaign! 

This competition is created and organized by the Advanced Concepts Team (ACT) with collaborations from Phi-Lab and the OPS-SAT spacecraft operations team.

Experts from both teams are available for interactions via the competition discussion page.

Created by the Advanced Concepts Team, Copyright © European Space Agency 2021