Self-supervised learning from unlabeled electron microscopy datasets facilitates efficient fine-tuning for various downstream tasks, demonstrating improved performance with limited annotated data.
