albumentations 1.4.3

An efficient library for image augmentation, providing extensive transformations to support machine learning and computer vision tasks.

Stars: 13344, Watchers: 13344, Forks: 1583, Open Issues: 431

The albumentations-team/albumentations repo was created 5 years ago and the last code push was 6 hours ago.
The project is extremely popular with a mindblowing 13344 github stars!

How to Install albumentations

You can install albumentations using pip

pip install albumentations

or add it to a project with poetry

poetry add albumentations

Package Details

Author

Vladimir I. Iglovikov, Mikhail Druzhinin, Alex Parinov, Alexander Buslaev, Eugene Khvedchenya

License

MIT

Homepage

https://albumentations.ai

PyPi:

https://pypi.org/project/albumentations/

GitHub Repo:

https://github.com/albumentations-team/albumentations

Classifiers

• Scientific/Engineering/Artificial Intelligence
• Scientific/Engineering/Image Processing
• Software Development/Libraries
• Software Development/Libraries/Python Modules

Related Packages

Errors

A list of common albumentations errors.

Code Examples

Here are some albumentations code examples and snippets.

GitHub Issues

The albumentations package has 431 open issues on GitHub

• is there a function for keypoints and bounding box corresponding removal
• Several frustrations when starting to work with albumentations on image-to-image problems
• Fix safe rotate targets
• Docs: fix inconsistent description for scale_limit in RandomScale and ShiftScaleRotate.
• Add keep ratio flag to Affine transform
• The mask information was incorrectly converted
• Fix bug in RingingOvershoot transform
• Bug in RingingOvershoot augmentation. "RuntimeWarning: invalid value encountered in true_divide"
• Library failed to import with opencv 4.5.5
• Affine doesn't rotate bounding boxes the same as Rotate
• keypoints augmentation for invisible keypoints
• WIP Fix Affine wrong rotation angle
• InvalidArgumentError: All dimensions except 3 must match. Input 1 has shape [8 48 48 512] and doesn't match input 0 with shape [8 40 40 288].
• Unclear warning when using Perspective transform in ReplayMode
• Corrected off-by-one in random crop.

See more issues on GitHub