cellsam
CellSAM is a universal model for cell segmentation. CellSAM achieves state-of-the-art performance on segmentation across a variety of cellular targets (bacteria, tissue, yeast, cell culture, etc.) and imaging modalities (brightfield, fluorescence, phase, multiplexed, etc.).
See the CellSAM paper for details!
Model Details
A Vision Transformer (ViT) model building on the detection and segmentation capabilities of the AnchorDETR and Segment Anything (SAM) models, respectively.
Model Description
- Developed by: Van Valen Lab
- Funded by: Funding details
- Model type: Vision transformer (ViT)
- License: Modified-apache2.0-noncommercial
Model Sources [optional]
- Repository: https://github.com/vanvalenlab/cellSAM
- Paper [optional]: https://www.nature.com/articles/s41592-025-02879-w
- Demo [optional]: https://vanvalenlab.github.io/cellSAM/auto_examples/index.html
Uses
Cell segmentation in microscope images.
Direct Use
The weights provided in this repository are suitable for cell segmentation across a variety of cellular targets (bacteria, tissue, yeast, cell culture, etc.) and imaging modalities (brightfield, fluorescence, phase, multiplexed, etc.).
Downstream Use
Cellsam is a generalist model designed to perform cell segmentation across all cell types and image modalities. However, it is possible to fine-tune CellSAM for specific applications. See the methods section of the paper for details.
Out-of-Scope Use
The model herein is provided subject to license terms prohibiting use for commercial applications.
Inquiries regarding commercial licensing should be directed to the Caltech Office of Technology Transfer.
How to Get Started with the Model
See the cellSAM tutorial and example gallery.
Evaluation
Evaluation protocols for reproducing the published results can be found in the documentation and source code repository.
Testing Data, Factors & Metrics
Testing Data
Evaluation data is available via users.deepcell.org.
Citation
BibTeX:
@article{marks2025cellsam, title={CellSAM: a foundation model for cell segmentation}, author={Marks, Markus and Israel, Uriah and Dilip, Rohit and Li, Qilin and Yu, Changhua and Laubscher, Emily and Iqbal, Ahamed and Pradhan, Elora and Ates, Ada and Abt, Martin and others}, journal={Nature Methods}, volume={22}, number={12}, pages={2585--2593}, year={2025}, publisher={Nature Publishing Group US New York} }
APA:
Marks, M., Israel, U., Dilip, R., Li, Q., Yu, C., Laubscher, E., ... & Van Valen, D. (2025). CellSAM: a foundation model for cell segmentation. Nature Methods, 22(12), 2585-2593.