AlbumentationsX is an actively developed image augmentation library applying consistent transforms to images, segmentation masks, bounding boxes, keypoints, and 3D data – a foundational tool for training more robust computer vision models.
The library is performance-optimized with reproducible public benchmarks comparing speed against competing computer vision libraries, built on a flexible NumPy-based interface supporting custom transforms and serializable pipelines across different training frameworks. Its comprehensive transform library is used across more than 40,000 public GitHub repositories.
AlbumentationsX targets researchers and academic institutions (cited in 2,270+ research papers), computer vision engineers and developers, and AI competition participants – having powered winning solutions in 77 competitions – across enterprise verticals including medical imaging, autonomous driving, geospatial analysis, life sciences, and industrial inspection. Open source under AGPL-3.0-only (`pip install albumentationsx`), with separately negotiated commercial licensing available through Albumentations, LLC. Affiliated with NumFOCUS.
