Project SMART wins first edition
of the AI Moonshot Challenge

The team led by the Center for Natural Resources and Environment (CERENA-IST) and which includes researchers from MIT will receive 500.000 Euro to develop a solution that combines satellite data and artificial intelligence to fight plastic pollution in the Oceans.
The AI Moonshot Challenge will return in 2021.

The SMART project, which brings together five Portuguese institutions and researchers from MIT Portugal, was the big winner of the first edition of the AI ​​Moonshot Challenge, which received applications involving institutions from 13 countries. The winner was announced this Thursday, December 3, 2020, during Web Summit 2020.

The AI ​​Moonshot Challenge, promoted by the Portuguese Space Agency, Portugal Space, in cooperation with the Portuguese Foundation for Science and Technology (FCT), Unbabel, European Space Agency (ESA), the Portuguese National Innovation Agency (ANI) and Web Summit, is an international competition that aims to find disruptive ideas combining satellite data and Artificial Intelligence for the detection of plastics in the Oceans.

From the ten applications submitted, the jury selected the proposal presented by the consortium led by CERENA-IST, which also includes the Underwater Systems and Technology Laboratory (FEUP), Associated Laboratory Center for Environmental and Sea Studies (CESAM), Hydrographic Institute (IH) and researchers from MIT Portugal, proposes to apply several advanced Artificial Intelligence concepts, with particular focus on “Physics Informed GANs and Deep Neural Networks”, and to combine them with satellite images to develop a solution to address the problem of plastics in the oceans.

In practical terms and simple language, the SMART project aims to combine machine learning with the known laws of physics to build models for predicting and simulating the accumulation of plastic in the Ocean. Using satellite data from the Copernicus program, the team intends to determine which frequencies are appropriate for the detection of plastic in bodies of water. This information will be complemented with advanced models that simulate the behaviour of the ocean to increase the probability of detecting plastics, a mission hampered by the size of the debris in relation to the resolution of the currently available satellite images.

The jury, co-chaired by biologist Carolina Sá, project manager at Portugal Space, and by Paolo Corradi, systems engineer at ESA, also valued the fact that the consortium’s partner network plans for the use of autonomous marine vehicles to validate the results and collect even more data.

From the applications submitted to the AI ​​Moonshot Challenge, which involve research institutes, universities, private companies and associations, in addition to state-owned companies, the jury also selected the proposals from the IMPLAST consortium (MARE-FCUL, LASIGE-FCUL, PML, MARE-UNova) and the ATLAS project (University of Coimbra, Remote Sensing Solutions, University of Aveiro, MARETEC-IST, AEBAM, GFZ) for the pitch finals during Web Summit. “It was a difficult selection given that all evaluators recognized the high quality of the proposals we received”, says Carolina Sá.

The Portugal Space project manager also mentions that the SMART project is the one that best responds to some of the parameters required in the evaluation of proposals. “The innovation factor in the use of satellite and Artificial Intelligence data, and the scientific component were the most prevalent criteria in the evaluation of the applications and we hope that the winning proposal can have a relevant contribution in solving this problem.

The president of the Portuguese Space Agency says that “the use of satellite data, namely from the Copernicus program, allows us to find large-scale solutions to some of the many challenges related to climate change”. “The AI ​​Moonshot Challenge and the answer it will provide to the problem of plastics in the Oceans are part of one of the main objectives of the Portuguese Space Agency – in this case of building a Digital Planet, a downstream digital platform capable of integrating multiple sources of data, including Space”, he adds.

For the minister of Science, Technology and Higher Education, Manuel Heitor, Space will be fundamental to combating climate change and preserving the Ocean. “The remote detection of plastics in the oceans based on the combination of satellite imagery with advanced artificial intelligence systems will be key for the sustainable future of the planet and its relationship with the climate. This award is part of the collective effort that is happening at the global level and will stimulate new research activities in Portugal ”, says Minister Manuel Heitor.

FCT also hopes that the AI Moonshot Challenge can have a relevant impact on the issue of climate change. “The Foundation for Science and Technology expresses its high expectations regarding the contribution of this competition for the solution of one of the main challenges facing humanity, allowing the development of national skills in the field of artificial intelligence and their application to the sustainability of the oceans ”, says José Paulo Esperança, vice president of the organization.

Launched during the 2019 edition of Web Summit, the AI ​​Moonshot Challenge is an international competition that seeks solutions combining Artificial Intelligence and diverse sources of satellite data, to solve complex problems in the context of climate change, in particular, and in its first edition, focused on the detection, location and monitoring of waste in the oceans.

“We had some of the best teams in the world competing to solve one of the biggest problems on our planet,” says Paulo Dimas, VP of Product Innovation at Unbabel, stressing that the company believes that “Artificial Intelligence will have a decisive role for the future of Humanity. Unbabel is an AI-first company that is contributing to this future”.

The detection of plastics through satellite data is an extremely complex task, given the distance the sensors are from objects, but also due to the spatial resolution of the data that is currently available. For example, the Copernicus Program has a spatial resolution of 10mx10m while a plastic bottle is considerably smaller. All of these factors contribute to creating “noise” in the satellite data, as well as making it easy to detect false positives (algae, organic debris, foam or even the crests of the waves), so the use of Artificial Intelligence becomes particularly relevant.

“Artificial Intelligence has the potential to be able to combine all this data and find patterns and make detections where we would not normally be able to,” says the Unbabel VP for Product Innovation. “The success of this edition is the result of a partnership led by Portugal Space that crosses the worlds of research and startups on the fantastic global stage that is the Web Summit. This partnership, sponsored by the Portuguese Foundation for Science and Technology (FCT) and with the support from the Portuguese Innovation Agency and ESA, will put Portugal in the leadership of a new domain – the intersection of Space and Artificial Intelligence technologies”, concludes Paulo Dimas.

The president of Portugal Space and the VP from Unbabel also announced that the AI ​​Moonshot Challenge will return in 2021 and that the challenge for the second edition will be announced soon.

Portugal Space
3 of December, 2020