medusa.geometry#

Module with geometry-related functionality.

For now only contains functions to compute vertex and triangle normals.

Module Contents#

medusa.geometry.compute_tri_normals(v, tris, normalize=True)[source]#

Computes triangle (surface/face) normals.

Parameters:

  • v (torch.tensor) – A float tensor with vertices of shape B (batch size) x V (vertices) x 3

  • tris (torch.tensor) – A long tensor with indices of shape T (triangles) x 3 (vertices per triangle)

  • normalize (bool) – Whether to normalize the normals (usually, you want to do this, but included here so it can be reused when computing vertex normals, which uses unnormalized triangle normals)

Returns:

fn – A float tensor with triangle normals of shape B (batch size) x T (triangles) x 3

Return type:

torch.tensor

medusa.geometry.compute_vertex_normals(v, tris)[source]#

Computes vertex normals in a vectorized way, based on the pytorch3d implementation.

Parameters:

  • v (torch.tensor) – A float tensor with vertices of shape B (batch size) x V (vertices) x 3

  • tris (torch.tensor) – A long tensor with indices of shape T (triangles) x 3 (vertices per triangle)

Returns:

vn – A float tensor with vertex normals of shape B (batch size) x V (vertices) x 3

Return type:

torch.tensor

medusa.geometry.apply_vertex_mask(name, **attrs)[source]#

Applies a vertex mask to a tensor of vertices.

Parameters:

  • v (torch.tensor) – A float tensor with vertices of shape B (batch size) x V (vertices) x 3

  • name (str) – Name of mask to apply

Returns:

v_masked – A float tensor with masked vertices of shape B (batch size) x V (vertices) x 3

Return type:

torch.tensor