Dream Team Epoch in top three percent of worldwide biomass AI competition

News - 07 February 2023 - TU Delft | Dream Hall

Dream Team Epoch of the TU Delft has achieved remarkable success by securing 21st place in the Artificial Intelligence (AI) BioMassters competition, ranking them among the top 3% of the 976 international competitors. The students have created an AI algorithm that accurately predicts the biomass of a forest region in Finland using satellite images. These predictions play a crucial role in research on the global carbon cycle and the implementation of regulations concerning forest conservation and controlled forest fires. With their algorithm, Team Epoch is making a significant impact towards creating a more sustainable world through the power of AI.

The competition organized by Mathworks aimed to find a more efficient way to calculate biomass. Currently, LiDAR data is utilised, which provides 3D information about the earth’s surface using laser technology. Although this method offers high precision data, it is time-consuming and costly due to the need for manual calibrations on the ground. To overcome these challenges, Epoch has developed an AI model that provides similar information, but without the use of the costly LiDAR method. The AI model uses images captured by the two Sentinel satellites of the European Space Agency, which have mapped the forest in Finland. By combining the data of these satellites, a single image is created that is as informative for research as the data collected through a LiDAR measurement. However, the Epoch algorithm that combines these images is faster and more accurate than manual methods, resulting in increased and improved biomass data for scientific research.

This was the second competition Epoch participated in this academic year. Previously, they placed ninth in the CityLearn Challenge. Team Manager Laura van der Hart expressed her pride in the team's growth and development, "I am incredibly proud of the team. It is clear that the engineers have grown in their role. They applied the lessons learned from the CityLearn Challenge to this competition with even more participants, a larger dataset, and greater complexity. This is truly a remarkable achievement!" AI Engineer Adrian Kuiper commented on the challenges of the competition, "A big challenge was generating one image from a huge amount of satellite data. The satellites gave us 180 images from which the algorithm had to take the most important information to produce that one image."  

Epoch focuses on participating in competitions that align with the UN Sustainable Development Goals, aiming to improve sustainability, safety, and health through the use of AI. Team Manager Laura says, "We believe that AI is a positive development with great potential, and therefore this technology should be applied as widely as possible." With two successful competitions, Epoch is now looking for their next challenge. There is still time for two more competitions before the end of the academic year, at which point a new group of students will form Epoch IV.