Mitral valve (MV) disease is a common pathologic problem occurring in approximately 2 % of the general population but climbing to 10 % in those over the age of 75. The preferred intervention for mitral regurgitation is valve repair, due to superior patient outcomes compared to those following valve replacement. Mitral valve interventions are technically challenging due to the functional and anatomical complexity of mitral pathologies. Repair must be tailored to the patient-specific anatomy and pathology, which requires considerable expert training and experience. Automatic segmentation of the mitral valve leaflets from 3D transesophageal echocardiography (TEE) may play an important role in treatment planning, as well as physical and computational modelling of patient-specific valve pathologies and potential repair approaches. This may have important implications in the drive towards personalized care and has the potential to impact clinical outcomes for those undergoing mitral valve interventions. To date, several groups have developed fully automatic segmentation approaches, achieving reported accuracy of 0.59 mm mean absolute surface distance. However, currently no public datasets are available to develop and compare these segmentation approaches using common data, thus it is difficult to draw conclusions based on reported accuracy alone. We aim to release the first public dataset of 3D TEE volumes of the mitral valve with annotations to enable the broader research community to address the mitral valve segmentation problem.