PianoCoRe: Combined and Refined Piano MIDI Dataset
:::info Stub — Full Engineering Breakdown Coming This paper was featured on Hugging Face Daily Papers on 2026-05-07 with 4 upvotes. A full breakdown with production viability rating, implementation notes, and honest limitations is being written. Subscribe to AI Letters → :::
| Authors | Ilya Borovik |
| Year | 2026 |
| HF Upvotes | 4 |
| arXiv | 2605.06627 |
| Download | |
| HF Page | View on Hugging Face |
Abstract
Symbolic music datasets with matched scores and performances are essential for many music information retrieval (MIR) tasks. Yet, existing resources often cover a narrow range of composers, lack performance variety, omit note-level alignments, or use inconsistent naming formats. This work presents PianoCoRe, a large-scale piano MIDI dataset that unifies and refines major open-source piano corpora. The dataset contains 250,046 performances of 5,625 pieces written by 483 composers, totaling 21,763 h of performed music. PianoCoRe is released in tiered subsets to support different applications: from large-scale analysis and pre-training (PianoCoRe-C and deduplicated PianoCoRe-B) to expressive performance modeling with note-level score alignment (PianoCoRe-A/A*). The note-aligned subset, PianoCoRe-A, provides the largest open-source collection of 157,207 performances aligned to 1,591 scores to date. In addition to the dataset, the contributions are: (1) a MIDI quality classifier for detecting corrupted and score-like transcriptions and (2) RAScoP, an alignment refinement pipeline that cleans temporal alignment errors and interpolates missing notes. The analysis shows that the refinement reduces temporal noise and eliminates tempo outliers. Moreover, an expressive performance rendering model trained on PianoCoRe demonstrates improved robustness to unseen pieces compared to models trained on raw or smaller datasets. PianoCoRe provides a ready-to-use foundation for the next generation of expressive piano performance research.
Engineering Breakdown
Plain English
This paper introduces PianoCoRe, a unified piano MIDI dataset combining and cleaning multiple open-source music corpora into a single resource containing 250,046 performances across 5,625 pieces by 483 composers (21,763 hours total). The dataset is released in tiered subsets optimized for different tasks: large-scale pre-training with 250K+ performances, and smaller curated subsets with note-level score alignments for expressive performance modeling.
Key Engineering Insight
The tiered release strategy—offering both massive deduplicated datasets for scaling and smaller carefully-aligned subsets for specialized tasks—directly addresses the practical engineering tradeoff between data volume and annotation quality that typically forces teams to choose one or the other.
Why It Matters for Engineers
Music generation and analysis models need symbolic training data, but existing datasets are fragmented across incompatible formats, inconsistent metadata, and varying quality standards. A unified, consistently-formatted dataset with clean metadata eliminates months of preprocessing work and enables direct comparison between different model architectures without data pipeline confounds.
Research Context
Previous MIR work relied on scattered, narrow datasets (limited composers, single performance per piece, no score-performance alignment). PianoCoRe unifies major open corpora with consistent naming, adds note-level alignment for expressive modeling tasks, and enables both large-scale pre-training (where scale matters more than perfection) and fine-grained performance analysis (where alignment quality is critical).
:::tip Subscribe Get weekly breakdowns of papers like this in AI Letters - the newsletter for engineers building production AI systems. :::