15k-star CV library stops dead; commercial fork takes over

What it does — Albumentations is a Python library for image augmentation in deep learning and computer vision. It provides a unified API to apply transforms to images, masks, bounding boxes, and keypoints across tasks like classification, segmentation, and object detection. The library claims benchmark wins as the fastest augmentation option and packs 70-plus transforms.

The interesting bit — The README is now a tombstone. As of June 2025 the project is officially unmaintained: no bug fixes, no compatibility patches, no new features. The team has moved all development to AlbumentationsX, a drop-in replacement that carries a dual AGPL-3.0/commercial license. If your project is MIT or Apache, the free version of the successor is legally off-limits.

Key highlights

• MIT-licensed forever, but frozen in time with no future updates.
• Supports RGB, grayscale, and multispectral images alongside masks and keypoints.
• Claims fastest augmentation library status in its own benchmarks.
• Part of the PyTorch ecosystem with integrations for TensorFlow and other frameworks.
• Used in production by Apple, Google, Meta, NVIDIA, Microsoft, and others (per the README’s logo wall).

Caveats

• No maintenance means it will likely rot against newer Python and PyTorch releases.
• The successor’s AGPL-3.0 license is explicitly incompatible with permissive open-source projects; commercial use requires a paid license.
• The “fastest” claims come from the project’s own benchmark page, not independent third-party verification shown in the README.

Verdict — Keep it if your pipeline is stable and you never need updates; otherwise, budget for a commercial license or plan a migration to a different library entirely. Researchers and hobbyists on permissive stacks should look elsewhere unless they want to open-source their entire project.