Computer Vision News - May 2023

32 Best Paper ISBI 2023 and made available for different tasks. However, these datasets often remain largely untapped due to the difficulties in performing transfer learning between them. Transfer learning in medical imaging is challenging due to the diversity of image modalities, organs, and cells and the scarcity of annotated data. Selecting an appropriate database for a given problem is still a significant hurdle. “ In the last few years, there has been a lot of focus on developing fancy new architectures to solve a specific task 0.5% better, ” Marcel tells us. “ However, this always assumes the availability of annotated data , which is not always practical in real-world applications. That’s Marcel Schilling is a doctoral student at the Institute for Automation and Applied Informatics at Karlsruhe Institute of Technology (KIT) in Germany. Miguel Molina-Moreno is a PhD student in the Multimedia Processing Group at the Charles III University of Madrid. They speak to us as the winners of the Best Paper Award at IEEE ISBI 2023 for their novel approach to tackling the problem of dataset similarity. Deep learning relies heavily on big datasets being created, annotated, AUTOMATED STYLE-AWARE SELECTION OF ANNOTATED PRE-TRAINING DATABASES IN BIOMEDICAL IMAGING B E S T PAPER I S B I 2 0 2 3 IEEE ISBI