MAESTRO: 200 Hours of Aligned Piano Performance Audio and MIDI

MAESTRO (MIDI and Audio Edited for Synchronous Tracks and Organization) v3.0.0 is a dataset of about 200 hours of virtuosic piano performances. It features fine alignment, approximately 3 milliseconds, between note labels and audio waveforms. The dataset was created by projectlosangeles and mirrored on Hugging Face as of August 3, 2025.

Use Cases

• Train automatic music transcription models based on aligned audio and MIDI data.
• Develop piano performance generation systems based on virtuosic performance examples.
• Research fine-grained audio-to-MIDI alignment techniques based on the 3 ms precision.
• Benchmark music information retrieval algorithms based on a large, synchronized dataset.

Strengths

• Approximately 200 hours of audio content provides substantial training material.
• Fine alignment of about 3 milliseconds between audio and note labels is a key feature.
• Focus on virtuosic piano performances offers a specific domain of high-quality data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and exact data structure are unknown, which may limit suitability assessment.

Provenance

Source

Hugging Face (projectlosangeles/maestro-v3.0.0), mirroring the original MAESTRO dataset.

Collection Method

Likely captured from live piano performances with synchronized recording equipment.

Time Range

null

Freshness

Last updated 2025-08-03 23:01:56; freshness should be verified.

Geography

null