Cost-efficient segmentation of electron microscopy images using active learning
Image analysis, particularly segmentation, in electron microscopy requires large amounts of manual labeling. Active learning aims to select samples smartly such that the required amount of labels can be reduced without affecting segmentation performance. This paper employs state-of-the-art concepts from active learning in classification and illustrates they can be generalized successfully to segmentation.
Joris Roels obtained his Master degree in Mathematical Computer Science and the PhD degree in Engineering from Ghent University in 1014 and 2019, respectively. He currently works as a postdoctoral researcher at VIB, where he is studying new machine learning methods for analyzing biological image data with minimal supervision.