Gert Kootstra is assistant professor in machine vision and robotics at the Farm Technology group of Wageningen University. His group focusses on developing, applying, and disseminating robotic technologies in the agro-food context. Gert received his PhD in artificial intelligence from the University of Groningen in 2009 and studied object detection and robotic grasping at the Royal Institute of Technology (KTH) in Stockholm from 2009-2012. He then specialized in the application of robotics in agriculture and food production at Wageningen University and Research. Gert coordinates FlexCRAFT, a large Dutch research program, with the aim to develop flexible agro-food robots. He is furthermore involved in several national and international research projects.
Agricultural robots encounter a lot of variation if you send them out in the field. Variation in the cultivars, growth stages, and illumination, for instance. In this presentation, I will discuss the challenges, show the importance to have sufficient variation in the training set, and propose an active-learning method to select the most relevant images for training.