alpine.utils package#

Submodules#

alpine.utils.checkers module#

alpine.utils.checkers.check_lossfn_types(loss_function)#
alpine.utils.checkers.check_opt_types(x)#

Check if the optimizer name is a string or a callable. If it is a string, check if it is in the list of supported optimizers. :param x: The optimizer name or callable. Must be a string or a torch.optim object. :type x: str or callable

alpine.utils.checkers.check_sch_types(x)#

Check if the scheduler is of the right type. :param x: learning rate scheduler, must be a torch.optim.lr_scheduler or partial object. :type x: callable

alpine.utils.checkers.wrap_signal_instance(x)#

alpine.utils.coords module#

alpine.utils.coords.get_coords2d(H, W)#

Get 2D coordinates for a grid size H x W.

Parameters:

  • H (int) – Height of field / image.

  • W (int) – Width of field / image.

Returns:

returns a tensor of shape (H*W, 2) with coordinates ranging from (-1, 1).

Return type:

torch.Tensor

alpine.utils.coords.get_coords3d(H, W, D)#

Get 3D coordinates for a grid size H x W x D.

Parameters:

  • H (int) – Height of field / image.

  • W (int) – Width of field / image.

  • D (int) – Depth of field / image.

Returns:

returns a tensor of shape (H*W*D, 3) with coordinates ranging from (-1, 1).

Return type:

torch.Tensor

alpine.utils.coords.get_coords3d_INT(H, W, D)#
alpine.utils.coords.get_coords_nd(*args, bounds=(-1, 1), indexing='ij')#

Get flattened coordinates for ND grid.

Parameters:

  • bounds (tuple, optional) – Bounds over the coordinates. Defaults to (-1,1).

  • indexing (str, optional) – Indexing scheme for meshgrid. Defaults to ‘ij’.

Returns:

Returns a tensor of shape (n, len(args)) with coordinates ranging from (-1, 1).

Return type:

_type_

alpine.utils.coords.get_coords_spatial(*args, bounds=(-1, 1), indexing='ij')#

Get spatial coordinates for ND grid.

Parameters:

  • bounds (tuple, optional) – Bounds over the coordinates. Defaults to (-1,1).

  • indexing (str, optional) – Indexing scheme for meshgrid. Defaults to ‘ij’.

Returns:

Returns a tensor of spatial coordinates across each dimension with coordinates ranging from (-1, 1).

Return type:

torch.Tensor

alpine.utils.functional module#

alpine.utils.functional.functional_model_call(model, params, coords)#
alpine.utils.functional.get_stacked_weights_and_bias_from_statedicts(params, N)#

alpine.utils.volutils module#

alpine.utils.volutils.march_and_save(occupancy, mcubes_thres, savename, smoothen=False)#

Convert volumetric occupancy cube to a 3D mesh

Inputs:

occupancy: (H, W, T) occupancy volume with values going from 0 to 1 mcubes_thres: Threshold for marching cubes algorithm savename: DAE file name to save smoothen: If True, the mesh is binarized, smoothened, and then the

marching cubes is applied

Outputs:

None

Module contents#

alpine.utils.normalize(x, mode='minmax')#
alpine.utils.normalize_nd(tensor, dim=None)#