Data Augmentation๏ƒ

This section covers data augmentation techniques for NIRS data.

Overview๏ƒ

Data augmentation is a powerful technique to improve model robustness, especially when working with limited training data. NIRS4ALL supports both sample-level and feature-level augmentation.

๐Ÿ”„ Augmentation Overview

Overview of augmentation methods and strategies.

Start Here

NIRS Data Augmentation in nirs4all
๐Ÿ“ˆ Sample Augmentation Guide

Complete guide to augmenting training data with synthetic samples, including balanced mode for imbalanced datasets.

Comprehensive

Sample Augmentation Guide
๐Ÿงช Synthetic Data

Generate synthetic NIRS spectra for testing and validation.

Testing

Synthetic NIRS Spectra Generator - Scientific Documentation

Types of Augmentation๏ƒ

Sample Augmentation๏ƒ

Create new training samples by applying transformations to existing spectra:

  • Noise injection

  • Baseline shifts

  • Wavelength perturbations

  • Mixup strategies

Feature Augmentation๏ƒ

Generate multiple views of the same data through different preprocessing:

  • Multiple preprocessing variants

  • Different spectral regions

  • Derivative combinations

Quick Example๏ƒ

from nirs4all.operators.augmentation import SampleAugmenter

pipeline = [
    {"sample_augmentation": SampleAugmenter(
        noise_level=0.01,
        n_augmented=3
    )},
    # ... rest of pipeline
]

See Also๏ƒ