PIAST: A Multimodal Piano Dataset with Audio, Symbolic and Text

TL;DR

PIAST is a comprehensive multimodal piano dataset with audio, symbolic, and text annotations, designed to advance Music Information Retrieval research by providing rich, annotated piano music data.

Contribution

The paper introduces PIAST, a new large-scale multimodal piano dataset with expert annotations, filling a critical gap in MIR resources for piano music.

Findings

Baseline results for music tagging and retrieval tasks.

Demonstrates dataset's potential for MIR research.

Includes both YouTube and expert-annotated subsets.

Abstract

While piano music has become a significant area of study in Music Information Retrieval (MIR), there is a notable lack of datasets for piano solo music with text labels. To address this gap, we present PIAST (PIano dataset with Audio, Symbolic, and Text), a piano music dataset. Utilizing a piano-specific taxonomy of semantic tags, we collected 9,673 tracks from YouTube and added human annotations for 2,023 tracks by music experts, resulting in two subsets: PIAST-YT and PIAST-AT. Both include audio, text, tag annotations, and transcribed MIDI utilizing state-of-the-art piano transcription and beat tracking models. Among many possible tasks with the multi-modal dataset, we conduct music tagging and retrieval using both audio and MIDI data and report baseline performances to demonstrate its potential as a valuable resource for MIR research.