Overview
Augmentor is a Python library designed to perform image augmentation for machine learning tasks. It provides a standalone, platform-independent solution to expand datasets, particularly for neural networks and deep learning models. The library employs a stochastic pipeline approach, where augmentation operations are applied probabilistically. Users can create pipelines with rotations, zooms, flips, and distortions. It supports multi-threading for faster processing and allows for augmentation of ground truth data in parallel with original images, suitable for tasks like image segmentation. Augmentor can be integrated with Keras and PyTorch, offering generators for on-the-fly data augmentation during training.