As our community grows we organize more and more events for our community. At the Helmholtz AI conference 2022 we could introduce two exciting challenges: the AI-Hero Hackathon on Energy Efficient AI and the Helmholtz AI CountMeIn Challenge.
As our community grows we organize more and more events for our community. At our Helmholtz AI conference we could introduce two exciting challenges. Take a deeper look at it here:
AI-Hero Hackathon on Energy Efficient AI
Markus Götz (KIT) presented the AI-HERO Hackathon on Energy Efficient AI in February 2022 organized by Helmholtz AI, Helmholtz Information and Data Science Academy (HIDA), Helmholtz Imaging (HI) and Helmholtz Metadata Collaboration (HMC), respectively DKFZ and KIT.
The hackathon took place from 1 February to 3 February 2022. This event featured two challenges: a use-case from the research field health and a use-case from the research field energy. Each task was tackled by six teams with three members each. The main focus of the tasks was not a development of a model that achieves higher accuracies, but was rather finding and optimizing a model and its training to minimize energy consumption while still maintaining a target accuracy. The winners in the Energy challenge were the `Dynamic Ants’ (T. Schanz, V. Zenchenko, S. Sharma, Hereon), the winners of the Health challenge the `Red Warriors` (M. Baumgartner, T. Wald, G. Köhler, DKFZ).
Helmholtz AI CountMeIn Challenge
Andres Camero (DLR) and Stefan Kesselheim (FZJ) introduced the Helmholtz AI CountMeIn Challenge that took place right before our Helmholtz AI conference from 12 April to 23 May 2022. In this challenge, the participating teams had to solve a challenging data science problem from the field of remote sensing: A spatially resolved estimate of the population based on satellite images. The task of the challenge was to solve the problem up to a target accuracy with the smallest amount of resources possible. In two tracks, two of the most precious resources were addressed. The GoFast track - in which the task is to use as little time as possible - was won by Yan Ji (FZJ). In the track GoGreen that used as little energy as possible Alina Bazarova (FZJ) and Sabrina Narimene Benassou (FZJ)won. The winning contributions to both tracks presented their solutions at the our Helmholtz AI conference 2022.
Images: Left: Andrés Camero (DLR), Stefan Kesselheim (FZJ) and Yan Ji (FZJ). Right: Alina Bazarova (FZJ) and Sabrina Narimene Benassou (FZJ).
For both tracks, the HAICORE installations located at Karlsruhe Institute of Technology (KIT) and Forschungszentrum Jülich (FZJ) were used. The submission happened on the newly formed Helmholtz Data Challenges Platform.