raman-data

A unified Python library for accessing public Raman spectroscopy datasets. raman-data is the dataset layer of RamanBench, a large-scale benchmark for machine learning on Raman spectroscopy data. It provides a single API to discover, download, and load 77 datasets from diverse sources (Kaggle, HuggingFace, Zenodo, Figshare, GitHub) in a standardized, ML-ready format.

Installation

pip install raman-data

Kaggle datasets require API credentials. Follow the Kaggle API setup guide and place your kaggle.json in ~/.kaggle/.

Quick Start

from raman_data import raman_data, TASK_TYPE, APPLICATION_TYPE

# List all available datasets
raman_data()

# Filter by task type or application domain
raman_data(task_type=TASK_TYPE.Classification)
raman_data(application_type=APPLICATION_TYPE.Medical)
raman_data(task_type=TASK_TYPE.Regression, application_type=APPLICATION_TYPE.Biological)

# Load a dataset by ID
dataset = raman_data("bioprocess_substrates")

# Access data
X = dataset.spectra          # np.ndarray (n_samples × n_wavenumbers)
w = dataset.raman_shifts     # np.ndarray of wavenumber values in cm⁻¹
y = dataset.targets          # np.ndarray of labels or values
print(dataset.metadata)

# Convert to pandas DataFrame
df = dataset.to_dataframe()

See examples/demo.ipynb for a full walkthrough.

RamanDataset API

Every raman_data(dataset_id) call returns a RamanDataset object:

Attribute	Type	Description
spectra	np.ndarray	Intensity matrix (samples × wavenumbers)
raman_shifts	np.ndarray	Wavenumber axis in cm⁻¹
targets	np.ndarray	Labels (classification) or values (regression)
target_names	list	Target column names (regression) or class names (classification)
metadata	dict	Source, paper reference, description
name	str	Dataset identifier
task_type	TASK_TYPE	Classification or Regression
application_type	APPLICATION_TYPE	Medical, Biological, Chemical, or MaterialScience
n_spectra	int	Number of spectra
n_frequencies	int	Number of wavenumber points
n_classes	int | None	Number of classes (classification only)
target_range	tuple | None	(min, max) of target values (regression only)
min_shift / max_shift	float	Spectral range in cm⁻¹

Variable-length spectra: If spectra share the same wavenumber axis, spectra and raman_shifts are standard 2D/1D arrays. If the axis differs per sample, both are returned as object arrays of 1D arrays. For machine learning, interpolate or pad to a common grid as needed.

Available Datasets

77 datasets across Medical, Biological, Chemical, and MaterialScience domains (click to expand)

Dataset Name	Application	Task Type	Description
acetic_acid_species	Chemical	Regression	Raman spectra and composition data for titration experiments of various acids in aqueous solution. Includes acetic, citric, formic, itaconic, levulinic, oxalic, and succinic acids. Data for concentration monitoring and indirect hard modeling.
adenine_colloidal_gold	Chemical	Regression	Quantitative SERS spectra of adenine measured using colloidal gold substrates across 15 different European laboratories. Benchmarks model reproducibility and inter-instrumental variability.
adenine_colloidal_silver	Chemical	Regression	Quantitative SERS spectra of adenine measured using colloidal silver substrates across 15 different European laboratories. Benchmarks model reproducibility and inter-instrumental variability.
adenine_solid_gold	Chemical	Regression	Quantitative SERS spectra of adenine measured using solid gold substrates across 15 different European laboratories. Benchmarks model reproducibility and inter-instrumental variability.
adenine_solid_silver	Chemical	Regression	Quantitative SERS spectra of adenine measured using solid silver substrates across 15 different European laboratories. Benchmarks model reproducibility and inter-instrumental variability.
alzheimer	Medical	Classification	Raman spectra from dried saliva drops targeting Alzheimer's Disease (PD) vs. healthy controls. Reveals hidden trends in proteins, lipids, and saccharides for early detection of cognitive and motor impairment.
amino_acids_glycine	Chemical	Regression	Time-resolved (on-line) Raman spectra for Glycine elution using a vertical flow LC-Raman method. Features 785 nm excitation and 0.2s exposure frames to benchmark label-free analyte detection.
amino_acids_leucine	Chemical	Regression	Time-resolved (on-line) Raman spectra for Leucine elution using a vertical flow LC-Raman method. Features 785 nm excitation and 0.2s exposure frames to benchmark label-free analyte detection.
amino_acids_phenylalanine	Chemical	Regression	Time-resolved (on-line) Raman spectra for Phenylalanine elution using a vertical flow LC-Raman method. Features 785 nm excitation and 0.2s exposure frames to benchmark label-free analyte detection.
amino_acids_tryptophan	Chemical	Regression	Time-resolved (on-line) Raman spectra for Tryptophan elution using a vertical flow LC-Raman method. Features 785 nm excitation and 0.2s exposure frames to benchmark label-free analyte detection.
bacteria_identification	Medical	Classification	60,000 spectra from 30 clinically relevant bacterial and yeast isolates (including an MRSA/MSSA isogenic pair). Acquired with 633 nm illumination on gold-coated silica substrates with low SNR to simulate rapid clinical acquisition times.
biomolecules_reference	Biological	Classification	Reference Raman spectra (450–1800 cm⁻¹, 1 cm⁻¹ resolution) of ~140 pure biomolecules including amino acids, nucleotides, lipids, and sugars. Each spectrum is labelled by biomolecule name. Useful for spectral assignment and as a reference library for classification benchmarks.
bioprocess_analytes_anton_532	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with an Anton Paar 532 nm spectrometer. Part of an 8-spectrometer cross-instrument series.
bioprocess_analytes_anton_785	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with an Anton Paar 785 nm spectrometer. Part of an 8-spectrometer cross-instrument series.
bioprocess_analytes_kaiser	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with a Kaiser spectrometer. Part of an 8-spectrometer cross-instrument series.
bioprocess_analytes_metrohm	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with a Metrohm spectrometer. Part of an 8-spectrometer cross-instrument series.
bioprocess_analytes_mettler_toledo	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with a Mettler Toledo spectrometer. Part of an 8-spectrometer cross-instrument series.
bioprocess_analytes_tec5	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with a Tec5 spectrometer. Part of an 8-spectrometer cross-instrument series.
bioprocess_analytes_timegate	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with a Timegate spectrometer. Part of an 8-spectrometer cross-instrument series.
bioprocess_analytes_tornado	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with a Tornado spectrometer. Part of an 8-spectrometer cross-instrument series.
bioprocess_substrates	Biological	Regression	A benchmark dataset of 6,960 spectra featuring eight key metabolites (glucose, glycerol, acetate, etc.) sampled via a statistically independent uniform distribution. Designed to evaluate regression robustness against common bioprocess correlations.
cancer_cell_(cooh)2	Biological	Classification	SERS spectra of cancer cell metabolites collected on gold nanourchins functionalized with the (COOH)2 moiety. Designed to provide specificity toward specific proteins and lipids for cell line identification.
cancer_cell_cooh	Biological	Classification	SERS spectra of cancer cell metabolites collected on gold nanourchins functionalized with the COOH moiety. Designed to provide specificity toward specific proteins and lipids for cell line identification.
cancer_cell_nh2	Biological	Classification	SERS spectra of cancer cell metabolites collected on gold nanourchins functionalized with the NH2 moiety. Designed to provide specificity toward specific proteins and lipids for cell line identification.
chembl_molecules	Chemical	Regression	140k DFT-computed Raman spectra for ChEMBL drug-like molecules. Targets: HOMO-LUMO gap, HOMO/LUMO energies, isotropic polarizability, heat capacity, dipole moment.
citric_acid_species	Chemical	Regression	Raman spectra and composition data for titration experiments of various acids in aqueous solution.
comfile_stroke	Medical	Classification	SERS serum spectra for binary stroke vs. healthy-control classification. ~4,020 spectra across 723 wavenumber points (202.985–1999.92 cm⁻¹).
covid19_salvia	Medical	Classification	Non-invasive SARS-CoV-2 screening from dried saliva drops. ~25 spectral replicates per subject from 101 patients (positive, negative symptomatic, and healthy controls), 785 nm excitation.
covid19_serum	Medical	Classification	Blood serum Raman spectra from 10 COVID-19 positive and 10 negative patients validated by RT-PCR and ELISA.
deepr_denoising	Unknown	Denoising	Raman spectral denoising dataset from the DeepeR paper. Noisy input spectra paired with denoised targets.
deepr_super_resolution	Unknown	SuperResolution	Hyperspectral super-resolution dataset from the DeepeR paper. Low-resolution inputs paired with high-resolution targets.
diabetes_skin_ages	Medical	Classification	AGEs signatures in skin Raman spectra, acquired in vivo with a portable 785 nm spectrometer to distinguish diabetic patients from healthy controls.
diabetes_skin_ear_lobe	Medical	Classification	Ear lobe skin Raman spectra for diabetic vs. healthy classification using a portable 785 nm spectrometer.
diabetes_skin_inner_arm	Medical	Classification	Inner arm skin Raman spectra for diabetic vs. healthy classification using a portable 785 nm spectrometer.
diabetes_skin_thumbnail	Medical	Classification	Thumbnail skin Raman spectra for diabetic vs. healthy classification using a portable 785 nm spectrometer.
diabetes_skin_vein	Medical	Classification	Vein skin Raman spectra for diabetic vs. healthy classification using a portable 785 nm spectrometer.
ecoli_fermentation	Biological	Regression	Batch and fed-batch E. coli fermentation spectra. Supernatant measurements with a 785 nm spectrometer tracking glucose and acetate concentrations.
ecoli_metabolites	Biological	Regression	Raman spectra of glucose and sodium acetate mixtures measured with an automated liquid handling station for high-throughput E. coli fermentation monitoring.
ecoli_metabolites_dig4bio	Biological	Regression	Raman spectra of glucose, sodium acetate, and magnesium sulfate mixtures measured with an automated high-throughput system for E. coli fermentation monitoring.
flow_microgel_synthesis	Chemical	Regression	In-line Raman spectra from a tubular flow reactor during microgel synthesis, paired with DLS-measured particle sizes.
formic_acid_species	Chemical	Regression	Raman spectra and composition data for titration experiments of various acids in aqueous solution.
fuel_benchtop	Chemical	Regression	Raman spectra from 179 commercial gasoline samples recorded with a benchtop 1064 nm FT-Raman system. Targets: RON, MON, and oxygenated additive concentrations.
fuel_handheld	Chemical	Regression	Same 179 gasoline samples as fuel_benchtop, acquired with a handheld 785 nm spectrometer. Benchmarks model transferability across hardware.
hair_dyes_sers	Chemical	Classification	SERS spectra of commercial hair dye products acquired with a portable Raman spectrometer. Target: brand identity.
head_neck_cancer	Medical	Classification	Raman spectra of blood plasma and saliva from head and neck cancer patients and healthy controls. Target: cancer vs. control (binary).
ht_raman_bio_catalysis_axp	Biological	Regression	Raman spectra for real-time monitoring of biocatalytic reactions in Deep Eutectic Solvents (DES).
illicit_adulterants_ft_raman	Medical	Classification	FT-Raman spectra (1064 nm) of 11 pharmaceutically active adulterants in dietary supplements. Target: compound identity.
illicit_adulterants_sers	Medical	Classification	SERS spectra (785 nm) of 11 illicit adulterants in dietary supplements. Target: compound identity.
itaconic_acid_species	Chemical	Regression	Raman spectra and composition data for titration experiments of various acids in aqueous solution.
kaiser_ecoli_fermentation	Biological	Regression	E. coli fermentation Raman spectra collected with a Kaiser spectrometer.
kaiser_ecoli_fermentation_supernatant	Biological	Regression	E. coli fermentation supernatant Raman spectra collected with a Kaiser spectrometer.
levulinic_acid_species	Chemical	Regression	Raman spectra and composition data for titration experiments of various acids in aqueous solution.
microgel_size_lf_fingerprint	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: Linear Fit, range: fingerprint region.
microgel_size_lf_global	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: Linear Fit, range: global.
microgel_size_mm_lf_fingerprint	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: MinMax + Linear Fit, range: fingerprint region.
microgel_size_mm_lf_global	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: MinMax + Linear Fit, range: global.
microgel_size_mm_rb_fingerprint	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: MinMax + Rubber Band, range: fingerprint region.
microgel_size_mm_rb_global	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: MinMax + Rubber Band, range: global.
microgel_size_raw_fingerprint	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: Raw, range: fingerprint region.
microgel_size_raw_global	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: Raw, range: global.
microgel_size_rb_fingerprint	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: Rubber Band, range: fingerprint region.
microgel_size_rb_global	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: Rubber Band, range: global.
microgel_size_snv_lf_fingerprint	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: SNV + Linear Fit, range: fingerprint region.
microgel_size_snv_lf_global	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: SNV + Linear Fit, range: global.
microgel_size_snv_rb_fingerprint	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: SNV + Rubber Band, range: fingerprint region.
microgel_size_snv_rb_global	Chemical	Regression	Raman spectra of 235 microgel samples (208–483 nm diameter). Pretreatment: SNV + Rubber Band, range: global.
microgel_synthesis	Chemical	Regression	In-line Raman spectra monitoring microgel synthesis in a tubular flow reactor with a customized measurement cell.
microplastics_weathered	MaterialScience	Classification	Raman spectra of 167 virgin and UV-weathered microplastic particles across multiple polymer types (PE, PP, PS, PET, PVC, etc.). Target: polymer type.
mlrod	MaterialScience	Classification	500,000+ Raman spectra of rock-forming silicate, carbonate, and sulfate minerals under Mars-like low-SNR conditions, without spectral preprocessing.
organic_compounds_preprocess	Chemical	Classification	Preprocessed Raman spectra of organic compounds from multiple excitation sources. Designed for transfer learning and domain adaptation benchmarks.
organic_compounds_raw	Chemical	Classification	Raw Raman spectra of organic compounds from multiple excitation sources. Designed for transfer learning and domain adaptation benchmarks.
parkinson	Medical	Classification	Raman spectra from dried saliva drops for Parkinson's Disease vs. healthy control classification.
pharmaceutical_ingredients	Medical	Classification	3,510 Raman spectra from 32 chemical substances including organic solvents and API development reagents. Target: compound identity.
ralstonia_fermentations	Biological	Regression	P(HB-co-HHx) copolymer synthesis monitoring in Ralstonia eutropha batch cultivations, combining experimental and high-fidelity synthetic data.
rruff_mineral_preprocess	MaterialScience	Classification	Preprocessed Raman spectra of 1,000+ mineral species from the RRUFF database, measured under varying conditions (532 nm and 785 nm).
rruff_mineral_raw	MaterialScience	Classification	Raw Raman spectra of 1,000+ mineral species from the RRUFF database, measured under varying conditions (532 nm and 785 nm).
serum_alzheimer_disease	Medical	Classification	Serum SERS spectra for classifying Alzheimer's disease (AD), Mild Cognitive Impairment (MCI), and healthy controls.
serum_prostate_cancer	Medical	Classification	Serum SERS spectra for classifying Prostate Cancer (PCa), Benign Prostatic Hyperplasia (BPH), and healthy controls.
streptococcus_thermophilus_fermentation_kaiser	Biological	Regression	Offline Raman spectra of Streptococcus thermophilus batch cultivations using a Kaiser RXN1 spectrometer. Two 24-hour fermentation runs in shake flasks.
streptococcus_thermophilus_fermentation_timegate	Biological	Regression	Offline Time-Gated Raman spectra of Streptococcus thermophilus batch cultivations. Two 24-hour fermentation runs in shake flasks.
succinic_acid_species	Chemical	Regression	Raman spectra and composition data for titration experiments of various acids in aqueous solution.
sugar_mixtures_high_snr	Chemical	Regression	High-SNR sugar mixture Raman spectra (7,680 measurements at 0.5 s integration) for benchmarking quantification algorithms.
sugar_mixtures_low_snr	Chemical	Regression	Low-SNR sugar mixture Raman spectra (7,680 measurements at 0.5 s integration) for evaluating noise robustness.
synthetic_organic_pigments_baseline_corrected	MaterialScience	Regression	Baseline-corrected spectral library of ~300 synthetic organic pigments for art conservation identification.
synthetic_organic_pigments_raw	MaterialScience	Regression	Raw spectral library of ~300 synthetic organic pigments for art conservation identification.
tg_ecoli_fermentation	Biological	Regression	E. coli fermentation Raman spectra collected using Time-Gated Raman Spectroscopy.
tg_ecoli_fermentation_supernatant	Biological	Regression	E. coli fermentation supernatant Raman spectra collected using Time-Gated Raman Spectroscopy.
wheat_lines	Biological	Classification	Raman spectra of salt-stress-tolerant wheat mutant lines and commercial cultivars (785 nm). Target: mutant line vs. cultivar.
yeast_fermentation	Biological	Regression	Raman spectra from continuous ethanolic fermentation of sucrose using Saccharomyces cerevisiae immobilized in calcium alginate beads.

Contributing a Dataset

We welcome contributions of new Raman datasets, especially from underrepresented domains, novel instrumentation, or larger sample sizes. A dataset is eligible for inclusion if it:

• Contains real, experimentally acquired 1D Raman spectra (not synthetic or simulated)
• Is publicly available without access restrictions or upon-request-only policies
• Provides ground-truth labels (class labels for classification; continuous values for regression)
• Is accompanied by a citable reference (paper, report, or dataset DOI)

Step-by-step

1. Host your data publicly. Upload your dataset to HuggingFace Datasets, Zenodo, Figshare, or Kaggle. HuggingFace is preferred for its versioning and direct streaming support — see dataset_to_huggingface.md for a step-by-step guide.

2. Choose the right loader. Add your dataset entry to the corresponding loader file in raman_data/loaders/:

Source	Loader file
HuggingFace	HuggingFaceLoader.py
Zenodo	ZenodoLoader.py
Figshare	FigshareLoader.py
Kaggle	KaggleLoader.py
Other URL (ZIP/file)	ZipLoader.py

3. Add a dataset entry. Each loader contains a DATASETS list. Add an entry following the existing pattern:

RamanDatasetConfig(
    dataset_id="your_dataset_id",        # unique snake_case identifier
    name="Your Dataset Name",            # human-readable name
    short_name="Short Name",             # for tables and figures (≤ 30 chars)
    task_type=TASK_TYPE.Regression,      # or TASK_TYPE.Classification
    application_type=APPLICATION_TYPE.Biological,
    description="One-sentence description of the dataset and task.",
    source="https://doi.org/...",        # DOI or URL of the dataset
    paper="https://doi.org/...",         # DOI of the associated paper
    loader_fn=_load_your_dataset,        # function that returns (spectra, shifts, targets)
)

4. Implement the loader function. The loader function must return a tuple (spectra, raman_shifts, targets):

def _load_your_dataset(config: RamanDatasetConfig):
    # Download / read your data here
    spectra      = ...  # np.ndarray, shape (n_samples, n_wavenumbers)
    raman_shifts = ...  # np.ndarray, shape (n_wavenumbers,), values in cm⁻¹
    targets      = ...  # np.ndarray, shape (n_samples,) or (n_samples, n_targets)
    return spectra, raman_shifts, targets

5. Add tests and open a pull request. Add a test in tests/ that loads your dataset and checks basic properties (shape, dtype, value ranges). Then open a pull request — we will review it and, if it meets the inclusion criteria, merge it and include it in the next RamanBench release.

Releasing a New Version

Releases are automated via GitHub Actions. A version tag triggers the CI pipeline: tests run first, then the package is built and published to PyPI automatically.

# Ensure all changes are committed and pushed to main
git checkout main && git pull

# Tag and push (uses setuptools-scm for versioning)
git tag v1.2.3
git push origin v1.2.3

The tag must match v*.*.*. The CI workflow will run tests across Python 3.10–3.13, build the package, and publish to PyPI.

Citation

If you use raman-data in your research, please cite the RamanBench paper:

@article{koddenbrock2026ramanbench,
  title={{RamanBench}: A Large-Scale Benchmark for Machine Learning on Raman Spectroscopy},
  author={Koddenbrock, Mario and Lange, Christoph and others},
  year={2026}
}

License

This project is licensed under the MIT License — see LICENSE for details. The individual datasets retain their original licenses as specified in each dataset's metadata.