Pose estimation models implemented in Pytorch Lightning, supporting massively accelerated training on unlabeled videos using NVIDIA DALI. Models can be evaluated with TensorBoard, FiftyOne, and Streamlit.
As of June 2024, Lightning Pose is now published in Nature Methods!
Train a network on an example dataset and visualize the results in Google Colab.
Please see the Lightning Pose documentation for installation instructions and user guides. Note that the Lightning Pose package provides tools for training and evaluating models on already labeled data and unlabeled video clips.
We also offer a browser-based application that supports the full life cycle of a pose estimation project, from data annotation to model training to diagnostic visualizations.
The Lightning Pose team also actively develops the Ensemble Kalman Smoother (EKS), a simple and performant post-processor that works with any pose estimation package including Lightning Pose, DeepLabCut, and SLEAP.
Lightning Pose is primarily maintained by Dan Biderman (Columbia University) and Matt Whiteway (Columbia University).
Lightning Pose is under active development and we welcome community contributions. Whether you want to implement some of your own ideas or help out with our development roadmap, please get in touch with us on Discord (see contributing guidelines here).