Automated olfactory bulb segmentation on high resolutional T2-weighted MRI

•First publicly available deep learning pipeline to segment the olfactory bulbs (OBs) in sub-millimeter T2-weighted whole-brain MRI.•Rigorous validation in the Rhineland Study - an ongoing large population-based cohort study - in terms of segmentation accuracy, stability and reliability of volume estimates, as well as sensitivity to replicate known OB volume associations (e.g. age effects).•Good generalizability to an unseen heterogeneous independent dataset (the Human Connectome Project).•Robustness even for individuals without apparent OBs, as can be encountered in large cohort studies.