→ PROBA-V Super Resolution
Enhance the vegetation payload performances
In this competition you are given multiple images of each of 78 Earth locations and you are asked to develop an algorithm to fuse them together into a single one. The result will be a "super-resolved" image that is checked against a high resolution image taken from the same satellite, PROBA-V. The 'V' stands for Vegetation, which is the main focus of the on-board instruments. Can you enhance the vision of PROBA-V and help us advance the accuracy on monitoring earths vegetation growth?
- 15.10.2018 - Data released.
- 01.11.2018 - Submissions are open.
- 23.01.2019 - A typo in the scoring evaluation has been fixed ("max" was changed to "min")
- 08.03.2019 - Midterm Online Meetup, 16:00 MET! (details)
- 01.06.2019 - Submissions are closed.
- 03.06.2019 - The final results are online! The first ranked team and thus the winner of the PROBA-V Superresolution challenge is SuperPip. Congratulations!
- 04.07.2019 - A preprint of a paper highlighting the generation of the PROBA-V Super Resolution dataset and our in-house approach has been posted to arxiv: Super-Resolution of PROBA-V Images Using Convolutional Neural Networks.
- 04.07.2019 - Silver medal winner team rarefin wrote an amazing blog post about the competition and their own solution strategy.
- 11.07.2019 - The Kelvins day has been officially announced!
- 22.07.2019 - Team SuperPip have a preprint of their work available on arxiv: DeepSUM: Deep neural network for Super-resolution of Unregistered Multitemporal images. Their code is available on this GitHub repository.
- 22.07.2019 - Team rarefin made their code public on this GitHub repository.
- 07.08.2019 - Submission can be evaluated again in the PROBA-V Super Resolution post mortem
- 29.08.2019 - The official paper about the design of the PROBA-V Super Resolution challenge has been published in the Astrodynamics journal. If you publish about this competition and data please refer to this work and no longer to the arxiv version!