PairwiseProjectionsLoss

class lightning_pose.losses.losses.PairwiseProjectionsLoss

Bases: Loss

Penalize projections from each pair of cameras into 3D world space.

Attributes Summary

loss_name

Methods Summary

__call__(keypoints_targ_3d, keypoints_pred_3d)	Call self as a function.
compute_loss(targets, predictions)
remove_nans(loss)

Attributes Documentation

loss_name = 'supervised_pairwise_projections'

Methods Documentation

__call__(keypoints_targ_3d: Tensor, {'__torchtyping__': True, 'details': ('batch', 'num_keypoints', 3), 'cls_name': 'TensorType'}], keypoints_pred_3d: Tensor, {'__torchtyping__': True, 'details': ('batch', 'cam_pairs', 'num_keypoints', 3), 'cls_name': 'TensorType'}], stage: Literal['train', 'val', 'test'] | None = None, **kwargs) → tuple[~torch.Annotated[~torch.Tensor, {'__torchtyping__': True, 'details': ((),), 'cls_name': 'TensorType'}], list[dict]]

Call self as a function.

compute_loss(targets: Tensor, {'__torchtyping__': True, 'details': ('batch', 'num_keypoints', 3), 'cls_name': 'TensorType'}], predictions: Tensor, {'__torchtyping__': True, 'details': ('batch', 'cam_pairs', 'num_keypoints', 3), 'cls_name': 'TensorType'}]) → Tensor, {'__torchtyping__': True, 'details': ('batch', 'cam_pairs', 'num_keypoints',), 'cls_name': 'TensorType'}]

remove_nans(loss: Tensor, {'__torchtyping__': True, 'details': ('batch', 'cam_pairs', 'num_keypoints'), 'cls_name': 'TensorType'}]) → Tensor, {'__torchtyping__': True, 'details': ('valid_losses',), 'cls_name': 'TensorType'}]

__init__(log_weight: float = 0.0, **kwargs) → None

Initialize PairwiseProjectionsLoss.

Parameters:

log_weight – final weight in front of the loss term in the objective function is computed as 1.0 / (2.0 * exp(log_weight)).

__new__(**kwargs)