Jürgen Gall:
Challenges and Opportunities of Representation Learning

Prof. Dr. Juergen Gall is professor and head of the Computer Vision Group at the University of Bonn since 2013. After his Ph.D. in computer science from the Saarland University and the Max Planck Institute for Informatics, he was a postdoctoral researcher at the Computer Vision Laboratory, ETH Zurich, from 2009 until 2012 and senior research scientist at the Max Planck Institute for Intelligent Systems in Tübingen from 2012 until 2013. He received a grant for an independent Emmy Noether research group from the German Research Foundation (DFG) in 2013, the German Pattern Recognition Award of the German Association for Pattern Recognition (DAGM) in 2014, and an ERC Starting Grant in 2016. He is further spokesperson of the DFG funded research unit “Anticipating Human Behavior” (http://for2535.cv-uni-bonn.de) and PI of the Cluster of Excellence “PhenoRob – Robotics and Phenotyping for Sustainable Crop Production” (http://www.phenorob.de).

Deep neural networks are used for many applications in the context of sustainable crop production. Examples are recognizing crops from RGB images, segmenting leaves or fruits, detecting plant diseases, or recognizing nutrient deficiencies. Deep neural networks, however, are not trained from scratch. Instead, pre-trained networks are commonly used and adapted to the data that has been collected for the so-called downstream task like recognizing nutrient deficiencies from RGB images.

The pre-training is also called representation learning and it is performed on very large datasets for image classification or video clip classification. The large-scale datasets, which are used for pre-training, however, are not related to the tasks that are relevant for crop production. In this talk, I will therefore give a brief introduction to representation learning and discuss the challenges and opportunities of representation learning in the context of crop production.