xnn.cnn.models.spherical.SO3Convolution#

class xnn.cnn.models.spherical.SO3Convolution(n_in, n_out, b_in, b_out, points)[source]#

Bases: Module

SO(3) correlation layer (eq 6, computed as the matrix product f^l psi^l+ of Appendix D).

Parameters:

  • n_in (int) – Input and output channels.

  • n_out (int) – Input and output channels.

  • b_in (int) – Bandwidth of the input and output SO(3) grids (b_out <= b_in).

  • b_out (int) – Bandwidth of the input and output SO(3) grids (b_out <= b_in).

  • points (Tensor) – Filter support, (n_points, 3) triples (beta, alpha, gamma).

filter_spectrum()[source]#

psi^l_{mn} of every channel pair, (n_in, n_out, b_out, 2 b_out - 1, 2 b_out - 1).

Return type:

Tensor

forward(x)[source]#

(B, n_in, 2 b_in, 2 b_in, 2 b_in) -> (B, n_out, 2 b_out, 2 b_out, 2 b_out).

Parameters:

x (Tensor)

Return type:

Tensor

extra_repr()[source]#

Set the extra representation of the module.

To print customized extra information, you should re-implement this method in your own modules. Both single-line and multi-line strings are acceptable.

Return type:

str