Abstract: Sound event localization and detection (SELD) requires simultaneous recognition of events and estimation of their spatial positions. Existing output representations have notable limitations.