Tuning A Distance-Prediction-Based Cell Segmentation

The virtually error-free segmentation of densely packed cells is still a challenging task. In 2020, we successfully participated as team KIT-Sch-GE (1) in the 5th edition of the ISBI Cell Tracking Challenge (CTC). By applying small adjustments in the training data augmentation, on the optimization setting and on the training data representation, we improved our deep learning-based segmentation further. Currently, we rank as team KIT-Sch-GE (2) several times within the top 3 on diverse cell data sets.