Albumentations 2025-07-03

[1] Albumentations has passed 15k stars on GitHub and reached 94M+ total downloads (94M+ total downloads).

[2] Albumentations is no longer maintained.

What does this mean?

The library has been forked into a new project: AlbumentationsX.

The corresponding Python package is: albumentationsx

It is a drop-in replacement for the original albumentations:

pip uninstall albumentations
pip install albumentationsx

Everything else stays the same:

import albumentations as A

transform = A.Compose([...])

Why?

The current situation

Over the past year, I’ve been the primary (and mostly only) maintainer of Albumentations.
The original team has moved on, but usage and support needs have only increased:

• Weekly issues, questions, and feature requests

• Major companies using it in production

• Sponsorships cover only 2.5% of my living expenses and 0.039% (0.0039) of my total compensation as a full time employee in my last job

Companies generally don’t donate — but they do purchase licenses.
The MIT model did not make long-term work sustainable.

📜 AlbumentationsX uses a dual license: AGPL / Commercial

Your options:

• ✅ Open-source projects using AGPL (or similarly restrictive licenses): You can keep using AlbumentationsX under AGPL.

• 💼 Companies have 3 paths:

  1. Open-source your code (to comply with AGPL)

  2. Purchase a commercial license

  3. Continue using the original albumentations (MIT) — stable, but not maintained

  🔍 Comparison

  License	MIT	AGPL / Commercial
  Active maintenance	No	Yes
  New features	No	Yes
  Bug fixes	No	Yes
  Code changes needed	–	None

🧠 What is AGPL?

AGPL extends GPL to cover network services.

• If you run AGPL software as a service, you must release the source code

• If your project is not AGPL-compatible (MIT, Apache, BSD), you must get a commercial license

This model ensures that improvements benefit the community — or are paid for properly.

🔁 Why this change makes sense

Other projects (like Ultralytics YOLO) have successfully adopted dual licensing to:

• Fund full-time development

• Deliver features faster

• Ensure stability and support

• Build sustainable open-source tools

My goal is to do the same: work on AlbumentationsX full-time and keep it the best augmentation library available.

This model won’t work for everyone — but it enables a level of support and velocity that the MIT model could not.

Questions? Feedback? Discussion welcome.