EuclideanDistance

class myoverse.models.components.losses.EuclideanDistance(n_joints=20, n_dims=3)

Euclidean distance loss for 3D joint positions.

Computes the mean Euclidean distance between predicted and ground truth 3D joint positions. Expects input tensors to be reshaped to (batch, joints, xyz).

Parameters:

• n_joints (int) – Number of joints in the skeleton. Default is 20.

• n_dims (int) – Number of dimensions per joint (typically 3 for x, y, z). Default is 3.

Examples

>>> loss_fn = EuclideanDistance(n_joints=20)
>>> pred = torch.randn(32, 60)  # batch_size=32, 20 joints * 3 dims
>>> target = torch.randn(32, 60)
>>> loss = loss_fn(pred, target)

Initialize internal Module state, shared by both nn.Module and ScriptModule.

Methods

__init__([n_joints, n_dims])	Initialize internal Module state, shared by both nn.Module and ScriptModule.
forward(prediction, ground_truth)	Compute the mean Euclidean distance loss.

forward(prediction, ground_truth)

Compute the mean Euclidean distance loss.

Parameters:

• prediction (torch.Tensor) – Predicted joint positions, shape (batch, n_joints * n_dims).

• ground_truth (torch.Tensor) – Ground truth joint positions, shape (batch, n_joints * n_dims).

Returns:

Scalar loss value.

Return type:

torch.Tensor