MulTTiPop: A Multitrack Transcription Dataset for Pop Music

2026-07-09

AI Summary

This study addresses a critical limitation in automatic music transcription research—the scarcity of high-quality, multi-track aligned datasets pairing popular music audio with MIDI. To bridge this gap, the authors construct a benchmark dataset comprising 572 pop songs (totaling 3.5 hours), which systematically integrates cross-era, multi-genre pop music audio with multi-track MIDI for the first time. High-precision temporal alignment is achieved through metadata matching, manual anchor point placement, audio beat tracking, and MIDI time warping. Evaluation of state-of-the-art transcription models on this dataset reveals a best onset F1 score of only 38%, underscoring substantial room for improvement in current methods while establishing a much-needed high-quality benchmark for the field.

Abstract

We present MulTTiPop, a dataset of pop music segments and their associated multitrack MIDI recordings for the evaluation of automatic music transcription models. MulTTiPop contains 572 segments of popular music totaling 3.5 hours of audio, and contains songs from diverse genres and decades from the 1930s to 2000s. To collect this dataset, we perform metadata-based matching on song segments from the Lakh MIDI and TheoryTab datasets, manually identify an anchor beat between the audio and MIDI, then use beat tracking on the audio and warp the MIDI to match its tempo and timing. We evaluate state-of-the-art automatic music transcription models on MulTTiPop and find substantial room for improvement, with the best model achieving 38% Onset F1. More details and sound examples of MulTTiPop are available at https://gclef-cmu.org/multtipop.

Problem

• automatic music transcription
• multitrack MIDI
• pop music dataset
• music transcription evaluation

Innovation

• multitrack transcription
• MIDI-audio alignment
• beat tracking
• pop music dataset
• automatic music transcription