tidy3d.plugins.invdes.FilterProject

tidy3d.plugins.invdes.FilterProject#

class FilterProject[source]#

Transformation involving convolution by a conic filter followed by a tanh projection.

Parameters:

radius (PositiveFloat) – [units = um]. Radius of the filter to convolve with supplied spatial data. Note: the corresponding feature size expressed in the device is typically sqrt(3) times smaller than the radius. For best results, it is recommended to make your radius about twice as large as the desired feature size. Note: the radius value is often only approximately related to the final feature sizes. It is useful to apply a FilterProject transformation to ‘encourage’ larger feature sizes, but we ultimately recommend creating a ErosionDilationPenalty to the DesignRegion.penalties if you have strict fabrication constraints.
beta (float = 1.0) – Steepness of the binarization, higher means more sharp transition at the expense of gradient accuracy and ease of optimization.
eta (float = 0.5) – Halfway point in projection function.
strict_binarize (bool = False) – If False, the binarization is still continuous between min and max. If True, the values are snapped to the min and max values after projection.

Notes

Attributes

Methods

evaluate(spatial_data, design_region_dl)

Evaluate this transformation on spatial data, given some grid size in the region.

evaluate(spatial_data, design_region_dl)[source]#: Evaluate this transformation on spatial data, given some grid size in the region.