PianoCoRe: Combined and Refined Piano MIDI Dataset
Researchers have introduced PianoCoRe, a comprehensive and refined dataset of piano MIDI performances and scores. This new dataset combines and improves upon existing open-source corpora, offering over 250,000 performances from 5,625 pieces by 483 composers. PianoCoRe includes tiered subsets for various applications, with a notable note-aligned collection for expressive performance modeling. The project also provides a MIDI quality classifier and an alignment refinement pipeline called RAScoP to enhance data accuracy. AI
IMPACT Provides a foundational dataset for advancing research in expressive piano performance modeling and music information retrieval.