RamanBench

A large-scale benchmark for machine learning on Raman spectroscopy data.

74 datasets · 163 prediction targets · 28 baseline models · 4 application domains

RamanBench provides a reproducible evaluation protocol and a curated collection of public Raman spectroscopy datasets spanning Material Science, Biological, Medical, and Chemical applications. Researchers can rank new models against 28 pre-evaluated baselines — from classical PLS to tabular foundation models and Raman-specific deep learning architectures — without re-running all experiments.

---

Ecosystem

raman-data   ──▶  raman-bench  ──▶  Live Leaderboard
(datasets)        (this package)     HuggingFace Space
PyPI / GitHub     PyPI / GitHub

Resource	Link
raman-data (dataset loader)	GitHub · PyPI
raman-bench (this package)	GitHub · PyPI
Live Leaderboard	huggingface.co/spaces/HTW-KI-Werkstatt/RamanBench
Paper	arXiv:2605.02003

---

Installation

Option 1 — Datasets + leaderboard (recommended starting point)

pip install raman-bench

This gives you:

• All 74 datasets with standardised train/test splits via raman-data
• Precomputed results for 28 baseline models (bundled CSVs, no internet needed)
• Leaderboard API — rank, plot, and compare against baselines
• Evaluation API — lb.evaluate_and_add(model) works with any sklearn-compatible model

You can use any ML library you already have installed — scikit-learn, LightGBM, XGBoost, PyTorch, JAX, or anything else — against a large-scale, curated data foundation without installing a single additional dependency.

Option 2 — With all built-in models

Adds all Raman-specific architectures and standalone tabular foundation models, all with a standard fit(X, y) / predict(X) interface:

pip install "raman-bench[models]"

This installs torch, tabpfn, pytabkit, tabdpt, sktime, and ramanspy on top of the core package. No AutoGluon required.

Option 3 — Full benchmark reproducibility (AutoGluon fork)

The paper's benchmark runs all models through AutoGluon's automated preprocessing and HPO pipeline. The fork addresses two limitations of standard AutoGluon 1.5:

1. Feature cap — AutoGluon caps tabular foundation models (TabPFN v2, TabICL, TabDPT, MITRA) at 500 features; Raman spectra typically have 500–4000 wavenumber points. The fork removes this cap.
2. TabICL v2 regression — AutoGluon 1.5 ships TabICL v1, which supports classification only. The fork upgrades to TabICL v2, adding regression support. This limitation is expected to be resolved in AutoGluon 1.6.

A patched fork incorporates both fixes.

git clone https://github.com/ml-lab-htw/RamanBench.git
cd RamanBench
pip install -r requirements-autogluon-fork.txt
pip install -e ".[models]"

The fork is only needed to reproduce the exact paper benchmark. Options 1 and 2 work with a standard pip install and give full access to all datasets, splits, and built-in models.

---

Quick Start

Load a dataset (Option 1 — core install only)

from raman_data import raman_data

ds = raman_data("amino_acids_glycine")
print(ds.spectra.shape)      # (n_samples, n_wavenumbers)
print(ds.targets.shape)      # (n_samples,)
print(ds.raman_shifts[:5])   # wavenumber axis in cm⁻¹

All 74 datasets are available this way. Each comes with a fixed train/test split so results are directly comparable to the precomputed baselines.

Evaluate your model against 28 baselines (Option 1)

Any scikit-learn–compatible estimator works:

from raman_bench import Leaderboard
from sklearn.cross_decomposition import PLSRegression

lb = Leaderboard.from_precomputed()   # loads bundled v0.1 results

# Evaluates on all 74 datasets (3 seeds) and inserts into the ranking
results = lb.evaluate_and_add(
    model_name="My-PLS-10",
    model=PLSRegression(n_components=10),
)
print(lb.rank())
lb.plot()

Bring any library — LightGBM, XGBoost, a PyTorch model, a JAX model — and it will be scored on the same protocol as the 28 precomputed baselines.

Explore the precomputed leaderboard (Option 1)

from raman_bench import Leaderboard

lb = Leaderboard.from_precomputed()
print(lb.rank())          # ranked DataFrame
lb.plot()                 # horizontal bar chart

Use a built-in Raman model directly

All built-in models expose a standard sklearn fit / predict API:

import numpy as np
from raman_bench.models.custom import DeepCNNModel, TabPFNModel, RocketModel

X = np.random.randn(200, 512).astype("float32")  # 200 spectra, 512 wavenumbers
y = np.random.randn(200)                          # regression targets

# Raman-specific deep learning model
model = DeepCNNModel(n_epochs=50)
model.fit(X, y)
predictions = model.predict(X)

# Tabular foundation model (no feature-count limit)
tfm = TabPFNModel()
tfm.fit(X, y)
predictions = tfm.predict(X)

Run the full benchmark pipeline (fork required)

# Pre-cache all dataset splits (optional, speeds up the run)
python scripts/prepare_datasets.py --config configs/benchmark_v0.1.json

# Run predictions → metrics
raman-bench run --config configs/benchmark_v0.1.json

# Run individual steps
raman-bench run --config configs/benchmark_v0.1.json --step predictions
raman-bench run --config configs/benchmark_v0.1.json --step metrics

Notebooks

Notebook	Description
01_quick_start.ipynb	Load a dataset, explore the precomputed leaderboard, plot rankings
02_benchmark_new_model.ipynb	Evaluate your own model and add it to the leaderboard
03_explore_results.ipynb	Deep dive into per-dataset and per-domain results
04_contribute_dataset.ipynb	Step-by-step guide to contributing a new dataset

---

Models

Paper baselines (28 models)

All results in the paper were produced through the AutoGluon pipeline (Option 3 install).

| Category | Models |

Category	Models
Classical spectroscopy	PLS, KNN, LR
Tree ensembles	GBM (LightGBM), XGB, CatBoost, RF, XT
Tabular deep learning	NN_TORCH, FastAI, RealMLP
Tabular foundation models	TabPFN v2, TabPFN v2.5, TabM, TabDPT, TabICL, MITRA
Time-series classifiers	ROCKET, Arsenal
Raman-specific DL	DeepCNN, RamanNet, SANet, RamanFormer, RamanTransformer, ReZeroNet, FC-ResNeXt, CoAtNet
AutoGluon ensemble	AUTOGLUON

Standalone sklearn wrappers (raman-bench[models])

raman-bench[models] provides sklearn-compatible (fit / predict) wrappers for many of the same algorithm families, usable directly without AutoGluon or the fork. These are not the exact pipeline configurations from the paper (no AutoGluon preprocessing or HPO), but they use the same underlying algorithms and are well-suited for building and evaluating new models.

Class	Algorithm	Requires
PLSModel	Partial Least Squares	—
DeepCNNModel	Raman-specific CNN	torch
RamanNetModel	Raman-specific CNN	torch
SANetModel	Spectral attention net	torch
RamanFormerModel	Raman transformer	torch
RamanTransformerModel	Raman transformer	torch
ReZeroNetModel	ReZero CNN	torch
FCResNeXtModel	FC-ResNeXt	torch
CoAtNetModel	Conv + attention	torch
RocketModel	ROCKET classifier	sktime
ArsenalModel	Arsenal classifier	sktime
TabPFNModel	TabPFN v2	tabpfn
RealMLPModel	RealMLP-TD	pytabkit
TabMModel	TabM-D	pytabkit
TabDPTModel	TabDPT	tabdpt

All classes support classification and regression and auto-detect the task from y. All package dependencies are included in raman-bench[models].

---

Benchmark Composition

Datasets

74 public Raman spectroscopy datasets from four application domains:

Domain	Datasets	Task	Sources
Chemical	37	Regression	Zenodo, HuggingFace
Medical	11	Classification	Kaggle, Zenodo
Biological	8	Regression	HuggingFace, Zenodo
Material Science	4	Classification	RRUFF, Zenodo

All datasets are accessible via pip install raman-data:

from raman_data import raman_data

dataset = raman_data("amino_acids_glycine")
X = dataset.spectra          # (n_samples, n_wavenumbers)
y = dataset.targets          # regression targets or class labels
w = dataset.raman_shifts     # wavenumber axis in cm⁻¹

Dataset catalog: raman-data on GitHub

---

Ranking Protocol

Models are evaluated under three complementary metrics:

Metric	Description
Elo	Pairwise win-rate Elo calibrated to RF = 1000 (200-round bootstrap)
Score	Normalised per-dataset score: best model = 1, median model = 0
Avg Rank	Average rank across all datasets and targets
Improvability	% gap to the best model, averaged across datasets

See the live leaderboard for interactive filtering by model category, task type, and dataset domain.

---

Repository Structure

RamanBench/
├── src/raman_bench/
│   ├── leaderboard.py          # Leaderboard + model evaluation API
│   ├── benchmark.py            # Dataset loading and cross-validation
│   ├── predictions.py          # Prediction generation (benchmark step 1)
│   ├── evaluation.py           # Metric computation (benchmark step 2)
│   ├── model.py                # AutoGluon pipeline wrapper (fork required)
│   ├── config.py               # JSON config loader
│   ├── models/custom/          # All built-in Raman models (sklearn API)
│   │   ├── base.py             #   BaseRamanEstimator (shared training loop)
│   │   ├── deepcnn.py          #   DeepCNNModel
│   │   ├── ramannet.py         #   RamanNetModel
│   │   ├── sanet.py            #   SANetModel
│   │   ├── ramanformer.py      #   RamanFormerModel
│   │   ├── ramantransformer.py #   RamanTransformerModel
│   │   ├── rezeronet.py        #   ReZeroNetModel
│   │   ├── fcresnext.py        #   FCResNeXtModel
│   │   ├── coatnet.py          #   CoAtNetModel
│   │   ├── pls.py              #   PLSModel
│   │   ├── sktime_models.py    #   RocketModel, ArsenalModel
│   │   └── tabular_foundation.py # TabPFNModel, RealMLPModel, TabMModel, TabDPTModel
│   └── preprocessing/
│       ├── mixin.py            #   RamanPreprocessingMixin (AutoGluon HPO)
│       └── wrapped_models.py   #   Prep_* classes + SklearnAutoGluonBridge
├── configs/                    # Benchmark configuration files
├── data/precomputed/           # Bundled v0.1 results
├── notebooks/                  # Example Jupyter notebooks
├── scripts/                    # CLI scripts
└── tests/                      # pytest test suite

Architecture: two paths, one set of model classes

Custom models are implemented once as plain scikit-learn BaseEstimator subclasses. The same classes are used in both usage modes:

  Custom model (e.g. DeepCNNModel)
  BaseEstimator — no AutoGluon dependency
  fit(X, y) / predict(X)
        │
        ├─── Standalone path (pip install "raman-bench[models]")
        │      CUSTOM_MODELS["DEEPCNN"] → DeepCNNModel().fit(X, y)
        │
        └─── AutoGluon pipeline path (fork required)
               SklearnAutoGluonBridge._fit() → DeepCNNModel(**params).fit(X_np, y_np)
               Prep_DEEPCNN(_RamanDLBase, _DeepCNNBridge)

SklearnAutoGluonBridge (in preprocessing/wrapped_models.py) is the only file that imports AutoGluon. All model source files are AutoGluon-free.

---

Contributing

We welcome contributions of new models and datasets!

Adding a New Model

The simplest way to add a model is to implement it as a scikit-learn–compatible estimator and submit a pull request. No AutoGluon knowledge is required.

1. Create src/raman_bench/models/custom/my_model.py:

import numpy as np
from sklearn.base import BaseEstimator

class MyModel(BaseEstimator):

    def __init__(self, n_components=10, lr=1e-3):
        self.n_components = n_components
        self.lr = lr

    def fit(self, X, y):
        # X: np.ndarray (n_samples, n_features)
        # y: np.ndarray — float → regression, int/str → classification
        ...
        return self

    def predict(self, X):
        ...  # return np.ndarray (n_samples,)

    def predict_proba(self, X):
        ...  # classification only, return (n_samples, n_classes)

For PyTorch-based models, inherit from BaseRamanEstimator in models/custom/base.py which provides a complete training loop with early stopping, cosine LR schedule, mixed-class augmentation, and batched inference.

2. Register in src/raman_bench/models/custom/__init__.py:

from raman_bench.models.custom.my_model import MyModel

CUSTOM_MODELS["MYMODEL"] = MyModel

3. Add tests in tests/models/test_my_model.py following the patterns in tests/models/test_sanet.py.

4. Open a pull request — CI will run the full test suite automatically.

See CONTRIBUTING.md for the full guide, including how to optionally wire your model into the AutoGluon benchmark pipeline for full reproducibility.

Adding a New Dataset

See CONTRIBUTING.md and NEW_DATASETS.md for detailed instructions and examples.

Quick summary:

1. Upload your dataset to HuggingFace Datasets or Zenodo under CC BY 4.0.
2. Add a loader to the raman-data package (open a PR there).
3. Open an issue here linking to the raman-data PR.

The live leaderboard also has a "How to Contribute" section with step-by-step instructions.

---

Citation

If you use RamanBench in your research, please cite:

@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}
}

---

License

MIT — see LICENSE.

Dataset licenses vary; see the dataset catalog or raman-data for per-dataset license information. Most datasets are released under CC BY 4.0.