Western Classical Music Recordings for Instrument Family Recognition

Michael Taenzer of the Fraunhofer Institute for Digital Media Technology created this data for reproducing results from an ISMIR conference paper. The repository contains audio data for experiment 2, which investigates CNN-based recognition of instrument families. Details on experiment 1 and the full dataset require contacting the authors.

Use Cases

• Training convolutional neural networks (CNNs) for instrument family classification based on audio recordings.
• Benchmarking patch-based versus file-based audio classification methods described in the paper.
• Reproducing the results of experiment 2 from the referenced ISMIR publication.

Strengths

• Data is associated with a peer-reviewed ISMIR conference paper, providing a research context.
• Created by a researcher at the Fraunhofer Institute for Digital Media Technology, suggesting an academic/technical origin.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total data size are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Michael Taenzer, Fraunhofer Institute for Digital Media Technology

Collection Method

Likely contains audio recordings or extracted features from Western classical music for machine learning experiments.