This zip file contains 1,592,753 heterogeneous, information-rich, non-redundant unlabeled 2D cellular EM (hence CEM1.5M) images, divided into 651 subdirectories (each directory being a unique vEM or EM image set). The image patches are mostly 224 x 224 pixels, however some are 512 x 512, and some are smaller. The raw image data was curated for deep learning largely following Conrad and Narayan, eLife 2021. https://elifesciences.org/articles/65894
