What is this?
Albumentations is a fast, flexible Python library for image augmentation used in training computer vision models, offering a huge set of pixel-level and spatial-level transforms for images, masks, bounding boxes, keypoints, and even 3D volumes.
Albumentations is a fast, flexible Python library for image augmentation used in training computer vision models, offering a huge set of pixel-level and spatial-level transforms for images, masks, bounding boxes, keypoints, and even 3D volumes.Is it a library or a standalone app?It is a Python library you import into your training pipeline, not a standalone application.
It is a Python library you import into your training pipeline, not a standalone application.How do I add it to a project?Install it with pip, then import it and build a composition of transforms to apply to your images, masks, bounding boxes, or keypoints.
Install it with pip, then import it and build a composition of transforms to apply to your images, masks, bounding boxes, or keypoints.Do I need PyTorch or TensorFlow to use it?The library itself is framework-agnostic (pure Python/NumPy-based transforms); it works with any image data regardless of whether you use PyTorch, TensorFlow, or another framework for training.
The library itself is framework-agnostic (pure Python/NumPy-based transforms); it works with any image data regardless of whether you use PyTorch, TensorFlow, or another framework for training.Is this project still maintained?No. The repository is no longer actively maintained as of June 2025, and no further bug fixes, features, or compatibility updates will be provided.
No. The repository is no longer actively maintained as of June 2025, and no further bug fixes, features, or compatibility updates will be provided.Who should actually use this versus the successor?Use original Albumentations only if your project already works fine with the current version and doesn't need fixes or new features; otherwise the README recommends migrating to AlbumentationsX for ongoing support.
Use original Albumentations only if your project already works fine with the current version and doesn't need fixes or new features; otherwise the README recommends migrating to AlbumentationsX for ongoing support.What kinds of data can transforms be applied to besides plain images?Transforms support images, segmentation masks, bounding boxes, keypoints, and volumetric 3D data (volumes and 3D masks), applying consistently across depth slices for 3D data.
Transforms support images, segmentation masks, bounding boxes, keypoints, and volumetric 3D data (volumes and 3D masks), applying consistently across depth slices for 3D data.How many augmentation types does it offer?It includes dozens of pixel-level transforms (like Blur, CLAHE, GaussNoise, RandomRain) and dozens of spatial-level transforms (like Affine, CoarseDropout, CropAndPad, ElasticTransform) that can affect images, masks, bounding boxes, and keypoints together.
It includes dozens of pixel-level transforms (like Blur, CLAHE, GaussNoise, RandomRain) and dozens of spatial-level transforms (like Affine, CoarseDropout, CropAndPad, ElasticTransform) that can affect images, masks, bounding boxes, and keypoints together.How do I migrate to AlbumentationsX if I need active support?The migration is designed to be a drop-in replacement: you simply pip install albumentationsx and your existing Albumentations code continues to work without changes.
The migration is designed to be a drop-in replacement: you simply pip install albumentationsx and your existing Albumentations code continues to work without changes.Can I use AlbumentationsX for free in a commercial MIT-licensed project?No — the AGPL-3.0 option is only free for AGPL-licensed projects and is incompatible with permissive licenses like MIT, Apache, or BSD, so commercial or permissively-licensed projects need a paid commercial license.
No — the AGPL-3.0 option is only free for AGPL-licensed projects and is incompatible with permissive licenses like MIT, Apache, or BSD, so commercial or permissively-licensed projects need a paid commercial license.What are similar projects?imgaug, labelImg, X-AnyLabeling, VisionHOPE, Detectron, detectron2, deep-learning-for-image-processing, kornia
imgaug, labelImg, X-AnyLabeling, VisionHOPE, Detectron, detectron2, deep-learning-for-image-processing, korniaWhich topics is it in?bounding-boxes, augmentation, detection, image-segmentation, segmentation, image-classification, object-detection, image-processing, deprecated, computer-vision, pytorch, deep-learning
bounding-boxes, augmentation, detection, image-segmentation, segmentation, image-classification, object-detection, image-processing, deprecated, computer-vision, pytorch, deep-learning