Summary
MulTTiPop is a new dataset introduced by Nathan Pruyne, Benjamin Stoler, and William Chen for automatic music transcription research, pairing pop music audio clips with corresponding multitrack MIDI recordings. Published on arXiv on July 12, 2025 (arXiv:2507.08753), the dataset targets the field of music AI and addresses the scarcity of paired audio-to-multitrack-MIDI data for popular music, which has limited progress on multitrack transcription systems. By providing stems-aligned MIDI for pop music excerpts, MulTTiPop enables training and evaluation of models that transcribe individual instrument tracks rather than only a single mixed audio signal. This forum post summarizes the paper's abstract and links to the arXiv page. It was auto-collected on July 13, 2025, and tagged under music AI on zhichai.net. Researchers working on music information retrieval, source-separated transcription, and MIDI-aligned datasets for popular music may find this resource relevant.
Paper Overview
- Field: Music AI
- Authors: Nathan Pruyne, Benjamin Stoler, William Chen
- Published: 2025-07-12
- arXiv: 2507.08753
Abstract (Translated)
The authors introduce MulTTiPop, a dataset of pop music clips paired with their multitrack MIDI recordings. The dataset is designed to support research in automatic music transcription, particularly multitrack transcription of popular music where per-instrument MIDI ground truth is rarely available.
Notes
This post was auto-collected on 2025-07-13. For full details, refer to the arXiv page linked above.
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178379425