U-Net basierte Koloniedetektion in Fluoreszenzbildern

Abstract Manual analysis of bacterial colonies in microbiology is time-consuming and error-prone. This study examines the suitability of U-Net models for the automated detection of colonies on fluorescent images. A particular advantage of these models is the pixel-precise segmentation, which enables detailed analyses. The model performed effectively with an F1 score of 0.93 on the validation data and a mask prediction time of just 0.27 seconds without the need for image preprocessing, even in the presence of artifacts. The implementation uses freely available libraries, which simplifies access.