CellMap Segmentation Challenge

The CellMap Segmentation Challenge white paper presents a comprehensive dataset of 289 annotated training crops derived from 22 distinct eFIB-SEM datasets, covering over 40 unique organelle and subcellular structure classes. This resource is designed to accelerate advancements in machine learning-based segmentation of electron microscopy data. The paper details the dataset’s biological diversity, preparation protocols, annotation standards, and associated metadata. It highlights rigorous quality control measures and the open-science principles that underpin the dataset’s public availability. This work serves as a foundational resource for developing, benchmarking, and enhancing image segmentation models, fostering discoveries in cellular architecture and advancing the broader field of computational microscopy.