project_3d_to_2d

lightning_pose.data.cameras.project_3d_to_2d(points_3d: Float[Tensor, 'batch num_keypoints 3'], intrinsics: Float[Tensor, 'batch num_views 3 3'], extrinsics: Float[Tensor, 'batch num_views 3 4'], dist: Float[Tensor, 'batch num_views num_params']) → Float[Tensor, 'batch num_views num_keypoints 2']

Project 3D keypoints to 2D using camera parameters.

Fully vectorized and differentiable implementation.

Parameters:

• points_3d – 3D points in world coordinates

• intrinsics – Camera intrinsic matrices (3x3)

• extrinsics – Camera extrinsic matrices (3x4)

• dist – Camera distortion parameters

Returns:

2D projected points for each camera view