Abstract: The VocalNotes dataset is a collection of audio and annotations for excerpts of vocal performances from five musical traditions - Japanese Minyo, Chinese Hebei Bangzi opera, Russian traditional singing, Alpine yodel and Jewish Romaniote chant. For each tradition the dataset contains: about 10 minutes of audio; documentation for the songs from which annotated fragments originate; f0, independent onset, offset and note pitch annotations created by two or three experts; The dataset was created as part of the VocalNotes project [1]. It is released under CC-BY-NC-SA license and can be accessed by filling out a request form.