SpectraKit

Python toolkit for spectral data processing: smoothing, baseline correction, normalization, scatter correction, derivatives, peak analysis, and more.

SpectraKit is a lightweight, pip-installable library for preprocessing and analyzing spectral data from IR, Raman, and NIR spectroscopy. It follows a functional design with NumPy arrays as the primary data type and requires only NumPy + SciPy as core dependencies.

Documentation | API Reference | Project Page | Examples

Installation

pip install pyspectrakit

Note: The PyPI distribution name is pyspectrakit (due to a naming conflict). The import name is simply import spectrakit.

Optional extras for additional functionality:

pip install pyspectrakit[io]         # HDF5 file support
pip install pyspectrakit[cli]        # Command-line interface
pip install pyspectrakit[baselines]  # pybaselines backend (200+ methods)
pip install pyspectrakit[fitting]    # lmfit peak fitting
pip install pyspectrakit[sklearn]    # scikit-learn integration
pip install pyspectrakit[plot]       # Plotting utilities
pip install pyspectrakit[widgets]    # Jupyter interactive viewer
pip install pyspectrakit[all]        # Everything above

Quick Start

import numpy as np
from spectrakit import smooth_savgol, baseline_als, normalize_snv

# Load your spectral data (N spectra, W wavelengths)
spectra = np.loadtxt("data.csv", delimiter=",")

# Process with individual functions
smoothed = smooth_savgol(spectra, window_length=11)
corrected = baseline_als(smoothed, lam=1e6, p=0.01)
normalized = normalize_snv(corrected)

All functions accept both single spectra (W,) and batches (N, W).

Pipeline

Chain steps for reproducibility:

from spectrakit.pipeline import Pipeline

pipe = Pipeline()
pipe.add("smooth", smooth_savgol, window_length=11)
pipe.add("baseline", baseline_als, lam=1e6)
pipe.add("normalize", normalize_snv)

processed = pipe.transform(spectra)

scikit-learn Integration

Use any SpectraKit function in an sklearn pipeline:

from sklearn.pipeline import Pipeline as SkPipeline
from sklearn.decomposition import PCA
from sklearn.svm import SVC
from spectrakit.sklearn import SpectralTransformer

pipe = SkPipeline([
    ("smooth", SpectralTransformer(smooth_savgol, window_length=11)),
    ("baseline", SpectralTransformer(baseline_als, lam=1e6)),
    ("normalize", SpectralTransformer(normalize_snv)),
    ("pca", PCA(n_components=10)),
    ("svm", SVC()),
])

pipe.fit(X_train, y_train)
predictions = pipe.predict(X_test)

Features

Smoothing

Baseline Correction

Normalization

Derivatives

Scatter Correction

Spectral Transforms

Operations

Peak Analysis

Similarity Metrics

I/O Formats

Optional Backends

Visualization

from spectrakit.plot import plot_spectrum, plot_comparison, plot_baseline

Requires pip install pyspectrakit[plot].

Jupyter Widget

Interactive spectrum viewer for Jupyter notebooks, powered by SpectraView:

from spectrakit.widgets import SpectrumViewer

viewer = SpectrumViewer()
viewer.set_spectrum(wavenumbers, intensities, label="ethanol")
viewer

Requires pip install pyspectrakit[widgets].

Spectrum Container

from spectrakit import Spectrum

spec = Spectrum(
    intensities=np.array([...]),       # (W,) or (N, W)
    wavenumbers=np.array([...]),       # (W,), optional
    metadata={"instrument": "Bruker"},
    source_format="jcamp",
    label="ethanol_ir",
)

CLI

pip install pyspectrakit[cli]

spectrakit info ethanol.dx
spectrakit convert ethanol.dx ethanol.h5

Examples

See the examples/ directory for Jupyter notebooks:

1. Quick Start — basic preprocessing workflow
2. Baseline Methods — comparing correction algorithms
3. Derivatives & Peaks — derivative analysis and peak finding
4. Scatter Correction — MSC vs EMSC vs SNV
5. sklearn Pipeline — classification with preprocessing

Development

git clone https://github.com/ktubhyam/spectrakit.git
cd spectrakit
pip install -e ".[all,dev]"
pytest

See CONTRIBUTING.md for guidelines.

Citation

If you use SpectraKit in your research, please cite:

@software{spectrakit,
  author = {Karthikeyan, Tubhyam},
  title = {SpectraKit: Python toolkit for spectral data processing},
  url = {https://github.com/ktubhyam/spectrakit},
  license = {MIT}
}

License

MIT