African contestants take top prizes at global satellite data labelling contest
Information from Radiant Earth Foundation
Radiant Earth Foundation has organised a data labelling contest as part of the Cloud Native Geospatial Outreach Day sponsored by Planet, Microsoft, and Azavea. The contest was designed as a crowdsourcing campaign to encourage the global community to contribute to open-access training data catalogues. Participants were asked to identify cloudy pixels in Sentinel-2 scenes, to help develop a large-scale accurate cloud detection training dataset.
The labelling contest was conducted on GroundWork, Azavea’s annotation tool designed for geospatial data. In the end, 231 users around the world signed up and labelled 75 645 tasks, which equates to about 2 million km2 of classified Sentinel-2 imagery.
Several scenes from Digital Earth Africa’s (DEA) Sentinel-2 catalogue were selected. DEA’s team converted all of the Sentinel-2 imagery across Africa to COG and hosted them on AWS. The resulting training dataset will be hosted on Radiant MLHub with a CC BY 4.0 license for public access.
The Top Labeler first prize has been awarded to Solomon Kica from Uganda and Jhomira Vanessa Loja Zumaeta from Peru. They each will receive a cash prize of $2000, plus an open 50 cm SkySat Image, tasked by each. Since their scores were incredibly close and stood out from the rest, both are awarded the top prize, and the second-place labeller prize thus removed.
Victor Ademoyero from Nigeria won the third place Labeler prize, with Frank Anderson from Tanzania selected as the Top Labeler from an African Country, and Babalola Oluwabukola Kemisola from Nigeria as the Top Women Labeler. They each have won a cash prize of $200. Mercy Akintola (Nigeria), Rami Soufli (Algeria), Erick Mnyali Tamba (Tanzania), Olanrewaju Michael (Nigeria), and Emmanuel Ogwal (Uganda) won the Next 5 Top Labeler prizes.