tfg.rendering.volumetric.absorption.render

Renders a voxel grid using the absorption-only model, as described in "Escaping Plato's Cave: 3D Shape From Adversarial Rendering" (Henzler 2019).

tfg.rendering.volumetric.absorption.render(
    voxels: type_alias.TensorLike,
    absorption_factor: tf.float32 = 0.1,
    cell_size: tf.float32 = 1.0,
    axis: int = 2,
    name: str = 'absorption_render'
) -> tf.Tensor

In the following, A1 to An are optional batch dimensions.

voxels A tensor of shape [A1, ..., An, Vx, Vy, Vz, Vd], where Vx, Vy, Vz are the dimensions of the voxel grid and Vd the dimension of the information stored in each voxel (e.g. 3 for RGB color).
absorption_factor A scalar representing the density of the volume.
cell_size A scalar representing the size of a cell.
axis An index to the projection axis (0 for X, 1 for Y or 2 for Z).
name A name for this op. Defaults to "absorption_render".

A tensor of shape [A1, ..., An, Vx, Vy, Vd] representing images of size (Vx,Vy).

ValueError If the shape of the input tensors are not supported.