LabelledGreenData4All: Successful First Stakeholder Workshop
The subject of our first LabelledGreenData4All stakeholder workshop on June 13th 2024 was “Machine Learning Models and data Annotation and their Data Requirements within the Environmental Sector”. A total of 26 participants representing a variety of organisations from a wide range of backgrounds, such as science, researching, and public administration, joined us for the occasion.
Watch the workshop (in German) here:
We started off the workshop with a brief project presentation by Cathleen Mitzschke of the Umweltbundesamt (Department Z 2.3 „Digitale Transformation und Beratungsstelle Green IT“) and wetransform’s Franziska Hochenegger. This followed a keynote speech on “Data as the new gold” by Kevin Kocon of IGD Fraunhofer. In his presentation, Kevin demonstrated which challenges researchers face when it comes to availability and evaluation of training data and annotated data for machine learning. He also presented possible approaches for dealing with limited training data, such as pseudo-labelling and transfer learning.
Stefan Klinger of the Anwendungslabor für Künstliche Intelligenz und Big Data am UBA then gave a brief overview of current use cases within the environmental sector. These real-world applications range from the use of satellite images to identify suitable locations for ground-mounted photovoltaic systems and windmills, to AI-supported analysis tools that help recognise the trade in illegal species on online trading platforms.
Thorsten Reitz, CEO of wetransform, led the subsequent discussion on the components of data availability, data processing and data infrastructure. The participants agreed that a lot of manual work is still required for the preparation of annotated data sets. The quality of both the data sets themselves and their coupled metadata, are of vital importance here. This is also the reason why the reuse of annotated datasets can be extremely difficult. Participants also identified willingness to share data, as well as uncertainties surrounding the associated terms of use and licencing systems as limiting factors.
The contributions and findings from the workshop will be used as a basis for further requirements and potential analyses of data annotations within LabelledGreenData4All. The exchange will be continued and deepened in further stakeholder workshops and interviews with experts in the further course of the research project.
Our next online expert workshop will be in English and take place on September 24, 2024 from 14:30-16:00 CET.
We will present the Forest Data Space and discuss your ideas, questions and requirements.