SpectraLab

Build, validate, and deploy predictive models from NIR spectroscopy data.
_{40+ algorithms • full chemometrics pipeline • one-click model export • runs 100% locally}

---

What is SpectraLab?

An educational, open-source web application for near-infrared (NIR) spectroscopy. Upload your spectral data, walk through a guided pipeline -- preprocessing, exploration, variable selection, calibration, validation -- then export a trained model and standalone prediction script ready to deploy on embedded devices, lab instruments, or production lines.

No cloud. No accounts. Everything runs on your machine.

---

Quick Start

git clone <repo-url>
cd spectral_processing

python -m venv venv
venv\Scripts\activate        # Windows
# source venv/bin/activate   # Linux / macOS

pip install -r requirements.txt
python run.py

Open http://localhost:5000 and start building models.

---

Pipeline Overview

 Upload         Preprocess       Explore          Select           Model            Validate         Export
 ──────         ──────────       ───────          ──────           ─────            ────────         ──────
  CSV     ──►   SNV, MSC,   ──►  PCA, HCA,  ──►  SPA, GA,   ──►  PLS, SVR,  ──►  K-Fold,    ──►   .pkl
  X & Y         SG, deriv,       ICA, SOM,       CARS, iPLS      CNN, SVM,       LOO, perm        script
                baseline         MCR-ALS          LASSO           RF, kNN         test, T²         config

Stage	Methods
Preprocessing	Savitzky-Golay • 1st/2nd derivatives • SNV • MSC • baseline correction (AsLS, poly, rubberband) • detrend • normalization • wavelet denoising • OSC • cosmic ray removal
Exploration	PCA • HCA • ICA • MCR-ALS • CLS • Self-Organizing Maps • PARAFAC / Tucker3
Variable Selection	SPA • iPLS • Genetic Algorithm • CARS • LASSO • Forward Stepwise • Correlation / MI filters
Regression	MLR • PCR • PLS • SVR • LWR • ANN/MLP • Ridge • LASSO • Elastic Net • Kernel PLS • 1D CNN
Classification	LDA • PLS-DA • OPLS-DA • SIMCA • kNN • SVM • Random Forest • Gradient Boosting • Logistic Regression • Naive Bayes • Mahalanobis
Validation	K-Fold • LOO • Venetian Blinds • Repeated K-Fold • Permutation test • Hotelling T² • Q-residuals
Metrics	RMSEP • R² • Bias • RPD • RER • Accuracy • Precision • Recall • F1 • Confusion Matrix

---

Model Export & Deployment

After training and validating, click Export to download a deployment package (.zip):

File	Purpose
regression_model.pkl	Trained model serialized with joblib
predict_regression.py	Self-contained prediction script -- preprocessing, variable selection, and inference are all embedded
model_config.json	Full config: preprocessing steps, wavelength indices, metrics
README.txt	Quick-start instructions

Run predictions anywhere

pip install numpy pandas scikit-learn joblib scipy
python predict_regression.py unknown_spectra.csv
# => writes unknown_spectra_predictions.csv

Deployment targets

• Embedded devices -- Raspberry Pi, Jetson Nano, industrial PCs
• Lab instruments -- NIR sensors with a Python runtime
• Production QC -- real-time or batch inline predictions
• Edge / cloud -- integrate into existing data pipelines
• Offline environments -- no internet required after export

---

Data Format

Spectral data and reference values are uploaded as separate CSV files.

Spectral data (X):

900,902,904,...,1700
0.453,0.461,0.472,...,0.321
0.412,0.418,0.425,...,0.298

• Column headers = wavelength values (numeric)
• Rows = samples

Reference values (Y):

moisture
12.5
8.3

• Single column, descriptive header
• One value per sample, matching row order

Sample Datasets

Included in data/sample/ for one-click loading:

File	Contents
nir_regression_spectra.csv	80 samples × 400 wavelengths
nir_regression_reference.csv	Moisture reference values
nir_classification_spectra.csv	90 samples × 400 wavelengths
nir_classification_reference.csv	3 material class labels

---

Dependencies

Category	Packages
Web	Flask, Flask-Session
Core	NumPy, SciPy, Pandas, scikit-learn, joblib
Visualization	Plotly, Matplotlib
Deep Learning	PyTorch
Specialty	PyBaselines, PyWavelets, TensorLy, MiniSom, DEAP

---

Project Structure

spectral_processing/
├── app/
│   ├── __init__.py              # Flask app factory
│   ├── config.py                # Configuration
│   ├── session_manager.py       # Server-side session state
│   ├── routes/
│   │   ├── main.py              # Upload, sample data, session
│   │   ├── preprocessing.py     # Preprocessing pipeline
│   │   ├── exploratory.py       # PCA, HCA, ICA, etc.
│   │   ├── variable_selection.py
│   │   ├── regression.py        # Quantitative calibration
│   │   ├── classification.py    # Supervised discrimination
│   │   ├── validation.py        # Cross-validation & diagnostics
│   │   └── export.py            # Model download & script gen
│   ├── processing/              # Algorithm implementations
│   │   ├── preprocessing.py
│   │   ├── exploratory.py
│   │   ├── variable_selection.py
│   │   ├── regression.py
│   │   ├── classification.py
│   │   ├── deep_learning.py
│   │   └── validation.py
│   ├── templates/               # Jinja2 HTML templates
│   └── static/                  # CSS & JS
├── data/sample/                 # Bundled sample datasets
├── requirements.txt
├── run.py                       # Entry point
└── README.md

---

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/my-feature)
3. Commit your changes (git commit -m "Add my feature")
4. Push to the branch (git push origin feature/my-feature)
5. Open a Pull Request

---

License

MIT -- use freely for education, research, and production.

---

_{Built with Flask • scikit-learn • PyTorch • Plotly
Designed for chemometrics education and real-world NIR deployment}