Multiview

Lightning Pose supports multi-camera setups, where the same scene is captured from different angles with different cameras.

We offer a multi-view transformer solution that processes all views simultaneously, learning cross-view correlations to improve performance.

Note

As of October 2025, multi-view Lightning Pose does not yet support context frames or unsupervised losses.

Contents:

• Organizing your data
• Patch masking and 3D loss
• Training and inference