DeepCell
A Python library for single-cell image analysis. Run TensorFlow models locally with GPU support for cell segmentation and time-lapse tracking. Apache 2.0.
DeepCell is an open-source Python library for researchers analyzing individual cells in biological images. It runs locally and uses TensorFlow to identify whole cells or nuclei and track cells through time-lapse movies. Its Apache 2.0 license allows users to adapt the library for their own research.
The library covers whole-cell and nuclear segmentation in 2D and 3D images, plus cell tracking in 2D time-lapse datasets. Its scope includes multiplexed tissue images and live-cell imaging movies, so it suits work on both static samples and changes over time.
Researchers can apply pretrained models or develop their own. Included models handle fluorescent nuclei and whole cells captured with phase or fluorescence imaging. Using those models requires image pixels to match the physical scale of the training data. Annotated datasets provide material for model development, including fluorescent nuclear movies and static whole-cell images.
DeepCell supports local GPU processing through TensorFlow and Docker, with CUDA required for the documented GPU setup. The library is separate from DeepCell Kiosk, which deploys workflows for large datasets in the cloud.
Related tools cover other parts of the image-analysis workflow: DeepCell Toolbox handles preprocessing and model-output cleanup, DeepCell Tracking creates cell lineages, and DeepCell Label supports annotation of high-dimensional biological images for training data.
The library code and pretrained assets have different terms. Downloading DeepCell models and training datasets requires an API key, and those assets are licensed for non-commercial academic use only.
