RamanBench Dataset Mirror

⚠️ RESEARCH MIRROR ONLY — All datasets are provided for research/educational purposes. Original copyrights remain with original authors. See Sources & Licenses below.

---

A unified mirror of 87 Raman spectroscopy datasets from the RamanBench benchmark. Wide-format Parquet files for fast, reliable access.

Quick Start

from raman_bench import RamanBenchmark

# Fast mirror access (default)
bench = RamanBenchmark(
    dataset_names_classification=["wheat_lines"],
    dataset_names_regression=["amino_acids_glycine"],
    use_mirror=True  # Uses this HF mirror
)

Or load directly:

from datasets import load_dataset
ds = load_dataset("HTW-KI-Werkstatt/RamanBench", "wheat_lines", split="train")
df = ds.to_pandas()

Citation & Attribution

If using this mirror, cite RamanBench:

@article{koddenbrock2026ramanbench,
  title={RamanBench: A Large-Scale Benchmark for Machine Learning on Raman Spectroscopy},
  author={Koddenbrock, Mario and Lange, Christoph and Legner, Robin and J{\"a}ger, Martin and K{\"o}gler, Martin and Bournazou, Mariano N Cruz and Neubauer, Peter and Biessmann, Felix and Rodner, Erik},
  journal={arXiv preprint arXiv:2605.02003},
  year={2026}
}

CRITICAL: You MUST ALSO cite the original dataset sources. Failure to do so violates IP rights.

Dataset Structure

• Parquet files: Wide-format spectral data
  • Columns: wavenumber floats (e.g., "200.5") + targets (single or multi-target)
  • Data types: float32 for spectra, float64/string for targets
• Metadata: metadata.json with target names and task type

Available Datasets

87 datasets: 21 classification + 66 regression

Classification: wheat_lines, bacteria_identification, cancer_cell_*, hair_dyes_sers, ...

Regression: amino_acids_*, bioprocess_analytes_*, ecoli_fermentation, ...

See RamanBench paper for full list.

Sources & Licenses

Source	License	Copyright Holder
Kaggle	Dataset Terms	Authors/Kaggle
Zenodo	CC-BY, CC-0, custom	Depositors
Figshare	CC-BY	Depositors
GitHub (HTW-KI-Werkstatt)	CC-BY or dataset-specific	Contributors
HuggingFace Hub	Varies	Contributors
RWTH Aachen / Custom	Institution-dependent	Research groups

Before using: Check original source for current license and terms.

Legal Notice

USAGE RIGHTS:

• ✅ Research and non-commercial use
• ✅ Must cite original authors AND RamanBench
• ✅ Must respect original dataset licenses
• ❌ No redistribution without original author permission
• ❌ No commercial use unless licenses permit

INTELLECTUAL PROPERTY:

• Original authors/institutions retain all copyrights
• This mirror is for research convenience only
• You do NOT own or control these datasets

LIABILITY: By using this mirror, you acknowledge responsibility for respecting IP rights and following original dataset licenses.

Resources

• RamanBench GitHub: https://github.com/ml-lab-htw/RamanBench
• raman_data Package: https://github.com/ml-lab-htw/raman_data
• Paper: https://arxiv.org/abs/2605.02003
• Leaderboard: https://huggingface.co/spaces/HTW-KI-Werkstatt/RamanBench

---

Last updated: 2026-06-01 | License: CC-BY-4.0