In this study, we constructed an open dataset, which contains computed tomography (CT) data on nearly 500 vertebrate microfossils. We propose a semantic segmentation method for CT fish microfossil data based on deep learning (DL). We expect that our proposed method could be applied to CT data on other fossils with good performance. We also believe the fast-accumulating CT data on vertebrate microfossils might become a source of information-rich datasets for deep learning.

Coccolithophores are an important group of phytoplankton that produce intricate skeletons of calcium carbonate. They contribute to the base of the marine food web and are important drivers of the global carbon cycle. Here, we describe novel coccolithophores and novel life cycle combinations detected by electron microscopy in samples collected in the Red Sea and the western Mediterranean. Our study advances our understanding of coccolithophore diversity and life cycle complexity.