Lightly
LightlyLightlySSL, published as the lightly package, is a PyTorch library for self-supervised pretraining of computer-vision backbones on unlabeled images. It implements about two dozen methods, among them SimCLR, MoCo, BYOL, DINO, DINOv2, MAE and iBOT, from shared losses, heads and transforms, and also trains the LeWM latent world model. Lightly, a spin-off from ETH Zurich, builds it.
Openness
5 medium confidence- license
- MIT (OSI)
- source
- public (github.com/lightly-ai/lightly)
- core features withheld
- no — LightlySSL is published whole; Lightly's paid products, LightlyTrain and LightlyStudio, are separate products and withhold nothing from it
The library is MIT-licensed and builds from the public repository, with nothing held back from it. What Lightly sells, LightlyTrain for pretraining and fine-tuning and LightlyStudio for curating and labeling data, are separate products listed apart from the open LightlySSL framework.
- https://docs.lightly.ai/self-supervised-learning/examples/models.html recorded 2026-11-12
LightlySSL 1.6.1 documentation navigation lists "On-Premise Lightly One Worker" beside the open library's API reference.
- https://docs.lightly.ai/self-supervised-learning/index.html recorded 2026-11-12
LightlySSL 1.6.1 documentation banner: "Interested in training your own vision foundation model? Check out LightlyTrain!", pointing pretraining users to the separate product.
- https://pypi.org/pypi/lightly/json recorded 2026-11-12
PyPI record for lightly 1.6.1, uploaded 2026-11-10, classifier "License :: OSI Approved :: MIT License", Github github.com/lightly-ai/lightly.
- https://raw.githubusercontent.com/lightly-ai/lightly/HEAD/LICENSE.txt recorded 2026-11-12
LICENSE.txt is the MIT License text, with the copyright line "Copyright (c) 2018 The Python Packaging Authority".
- https://raw.githubusercontent.com/lightly-ai/lightly/HEAD/README.md recorded 2026-11-12
README: "For a commercial version with more features, including Docker support and pretraining models for embedding, classification, detection, and segmentation tasks with a single command, please contact sales@lightly.ai."
- https://www.lightly.ai/lightly-train recorded 2026-11-12
Site navigation lists "Core Products": "LightlyStudio Curate, label, and manage your data" and "LightlyTrain Pretrain your vision models, no labels needed", and separately "Open Source Projects": "LightlySSL Self-supervised learning framework". The LightlyTrain page offers "Commercial License AGPL-3.0 License Research License".
Adoption
2 high confidenceMeasured on monthly PyPI downloads of the lightly package. The package also carries the client for Lightly's hosted platform, so part of the figure may come from platform users rather than from self-supervised training.
- https://pypistats.org/api/packages/lightly/recent recorded 2026-11-12
last_month 94,597 downloads of lightly
Capability
3 medium confidenceLightlySSL is a toolkit of self-supervised training methods with recipes to pretrain and compare backbones, as timm is a zoo of architectures with training scripts. It ships no model that predicts on a user's own task without training; choosing and running a method is left to the user.
- https://docs.lightly.ai/self-supervised-learning/examples/models.html recorded 2026-11-12
Models page: "Self-supervised methods with runnable examples in PyTorch, PyTorch Lightning and distributed PyTorch Lightning. Each one is built from the same losses, heads and transforms".
- https://raw.githubusercontent.com/lightly-ai/lightly/HEAD/README.md recorded 2026-11-12
README: "LightlySSL is a computer vision framework for self-supervised learning and latent world models"; models table from MoCo (2019) to Pixio (2025), and LeWM (2026) under world models.
Verified 2026-11-12, except some scores, still awaiting re-confirmation