Dataset Viewer
Auto-converted to Parquet Duplicate
The dataset viewer is not available for this split.
Server error while post-processing the rows. This occured on row 17. Please report the issue.
Error code:   RowsPostProcessingError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Drum & Percussion Kits

Quick Start

from datasets import load_dataset

# Stream to avoid downloading the entire dataset
ds = load_dataset("schismaudio/drum-percussion-kits", streaming=True)

# Or download locally
ds = load_dataset("schismaudio/drum-percussion-kits")

Dataset Description

Drum & Percussion Kits is a diverse collection of drum and percussion one-shot samples assembled for use in synthesis and data augmentation pipelines. It combines free samples from SampleSwap, organic one-shot recordings, and 850+ programmatically generated samples. The full dataset is approximately 1.2 GB, distributed across 4 tar archives.

The dataset is intended as a broad-coverage sample library for training drum transcription and synthesis models, providing timbral variety beyond what any single recorded kit can offer.

Dataset Structure

Data Fields

Field Type Description
audio Audio WAV one-shot sample
filename string Original filename
instrument string Drum or percussion instrument label
source string Provenance tag: sampleswap, organic, or generated

Data Splits

This dataset has no predefined splits. All samples are in the default train split.

Archives

The dataset is packaged as 4 tar archives. Each archive contains a subset of the sample library organized by instrument family.

Usage Examples

Load and inspect samples

from datasets import load_dataset

ds = load_dataset("schismaudio/drum-percussion-kits")

sample = ds["train"][0]
print(sample["filename"], sample["instrument"], sample["source"])
audio_array = sample["audio"]["array"]
sr = sample["audio"]["sampling_rate"]

Filter by provenance

from datasets import load_dataset

ds = load_dataset("schismaudio/drum-percussion-kits")

organic = ds["train"].filter(lambda x: x["source"] == "organic")
print(f"Organic samples: {len(organic)}")

Dataset Creation

Source Data

The collection draws from three sources:

  • SampleSwap: Free-to-use drum samples from the SampleSwap community library, covering a wide range of drum machine and acoustic kit sounds.
  • Organic one-shots: Individually recorded percussion hits captured in various acoustic environments.
  • Generated samples: 850+ samples synthesized programmatically to expand coverage of underrepresented timbres and velocities.

Annotations

Instrument labels and source tags were assigned during curation. No onset or velocity annotations are included — this is a one-shot sample library, not a transcription dataset.

Known Limitations

  • Mixed provenance: Samples originate from multiple sources with different recording conditions, bit depths, and sample rates. Normalization and quality may be inconsistent.
  • Generated samples: The 850+ generated samples may not exhibit the same acoustic realism as recorded samples. Models trained on this data should be evaluated on real-world audio.
  • No standardized splits: Without defined train/test splits, users should partition the data carefully to avoid data leakage.

Related Datasets

This dataset is part of the Drum Audio Datasets collection by schismaudio. Related datasets:

Citation

@misc{drumpercussionkits,
  title     = {Drum \& Percussion Kits},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.3994999},
  url       = {https://zenodo.org/record/3994999}
}

License

This dataset is released under the Creative Commons Attribution 4.0 International License (CC-BY 4.0).

You are free to share and adapt this dataset for any purpose, including commercial use, as long as you give appropriate credit to the original authors.

Downloads last month
32

Collection including schismaudio/drum-percussion-kits