LakoreAI/stgram-mfn-dcase2020-eval

DCASE 2020 Task 2 — Evaluation + Additional Training (STgram-MFN redistribution)

Redistribution of the DCASE 2020 Challenge Task 2 Additional Training Dataset (Zenodo 3727685) and Evaluation Dataset (Zenodo 3841772), merged into the same on-disk layout as the development set.

Original authors: Yuma Koizumi, Yohei Kawaguchi, Keisuke Imoto. License: CC BY-NC-SA 4.0 (non-commercial) — inherited from the source; attribute the original authors and keep any redistribution under the same terms.

Layout

<machine>/{train,test}/<label>_id_<XX>_<index>.wav

• <machine> ∈ fan, pump, slider, valve, ToyCar, ToyConveyor
• train/ — additional normal clips (used as extra training data)
• test/ — normal + anomaly (held-out evaluation)
• audio: mono, 16 kHz, 10 s (.wav, 16-bit PCM)

Labels

No separate label file — encoded in the path:

Field	Source	Example
machine type	parent directory	fan
machine id (0–7)	id_XX token	id_00
normal / anomaly	filename prefix	anomaly → 1

Statistics

• 6 machine types, 23,267 clips, ~7.3 GB
• train: additional normal · test: normal + anomaly

Intended use

• train/ clips are concatenated onto the dev train/ set as the reference's add_dirs (extra normal-only data) — config key add_root, e.g. --add_root data/raw_eval.
• test/ is the held-out evaluation set for the backbone-swap / INT8 quantization study.

Usage

from huggingface_hub import snapshot_download
snapshot_download(
    repo_id="LakoreAI/stgram-mfn-dcase2020-eval",
    repo_type="dataset", local_dir="data/raw_eval",
)

uv run python scripts/training/train.py --config configs/train_lean.yaml \
    --add_root data/raw_eval

Related

• Development set: LakoreAI/stgram-mfn-dcase2020-dev
• Sources: https://zenodo.org/records/3727685 · https://zenodo.org/records/3841772

Citation

@inproceedings{koizumi2020dcase,
  title     = {Description and Discussion on DCASE2020 Challenge Task 2:
               Unsupervised Anomalous Sound Detection for Machine Condition Monitoring},
  author    = {Koizumi, Yuma and Kawaguchi, Yohei and Imoto, Keisuke and others},
  booktitle = {DCASE Workshop},
  year      = {2020}
}