rtgs.gaussian

Gaussian struct in Taichi.

Module Contents

Classes

Gaussian

Gaussian struct in Taichi.

Functions

new_gaussian

Python scope Gaussian constructor to create a new Gaussian.

Data

BOUNDING_THRESHOLD

c_0

c_1

c_2

c_3

c_4

c_5

c_6

API

rtgs.gaussian.BOUNDING_THRESHOLD = 3
rtgs.gaussian.c_0 = 'sqrt(...)'
rtgs.gaussian.c_1 = 'sqrt(...)'
rtgs.gaussian.c_2 = 'sqrt(...)'
rtgs.gaussian.c_3 = 'sqrt(...)'
rtgs.gaussian.c_4 = 'sqrt(...)'
rtgs.gaussian.c_5 = 'sqrt(...)'
rtgs.gaussian.c_6 = 'sqrt(...)'
class rtgs.gaussian.Gaussian

Gaussian struct in Taichi.

Parameters:

  • position – gaussian center (mean).

  • rotation – gaussian rotation quaternion.

  • scale – gaussian scale.

position: taichi.math.vec3 = None
rotation: taichi.math.vec4 = None
scale: taichi.math.vec3 = None
color: taichi.math.vec3 = None
opacity: taichi.f32 = None
sh_10: taichi.math.vec3 = None
sh_11: taichi.math.vec3 = None
sh_12: taichi.math.vec3 = None
sh_20: taichi.math.vec3 = None
sh_21: taichi.math.vec3 = None
sh_22: taichi.math.vec3 = None
sh_23: taichi.math.vec3 = None
sh_24: taichi.math.vec3 = None
sh_30: taichi.math.vec3 = None
sh_31: taichi.math.vec3 = None
sh_32: taichi.math.vec3 = None
sh_33: taichi.math.vec3 = None
sh_34: taichi.math.vec3 = None
sh_35: taichi.math.vec3 = None
sh_36: taichi.math.vec3 = None
init(position=ti.math.vec3(0, 0, 0), rotation=ti.math.vec4(0, 0, 0, 1), scale=ti.math.vec3(1, 1, 1), color=ti.math.vec3(1, 0, 1), opacity=1)

Taichi scope Gaussian initialization method.

Parameters:

  • position (ti.math.vec3) – gaussian center (mean).

  • rotation (ti.math.vec4) – gaussian rotation quaternion.

  • scale (ti.math.vec3) – gaussian scale.

  • color (ti.math.vec3) – gaussian base color.

  • opacity (ti.f32) – gaussian opacity.

cov()

Get the covariance matrix of the gaussian using the rotation quaternion and scale vector.

Returns:

a 3x3 covariance matrix.

Return type:

ti.math.mat3

bounding_box()

Compute the axis-aligned bounding box (AABB) for the Gaussian considering rotation.

Returns:

bounding box for the Gaussian.

Return type:

Bound

eval_sh(dir)

Evaluate spherical harmonics.

Parameters:

dir (ti.math.vec3) – normalized view direction.

Returns:

spherical harmonics radiance.

Return type:

ti.math.vec3

eval(pos, dir)

Evaluate gaussian color.

Parameters:

  • pos (ti.math.vec3) – evaluate position.

  • dir (ti.math.vec3) – ray direction for SH color encoding.

Returns:

gaussian color and alpha at pos from direction dir.

Return type:

ti.math.vec4

hit(ray)

Ray-Gaussian intersection test. The algorithm will only test the intersection for t between ray.start and ray.end.

Parameters:

ray (Ray) – camera ray.

Returns:

two ray Gaussian intersections in increasing order. If there’s no solution, the result will be both ti.math.inf.

Return type:

ti.math.vec2

rtgs.gaussian.new_gaussian(position: taichi.math.vec3 = ti.math.vec3(0, 0, 0), rotation: taichi.math.vec4 = ti.math.vec4(0, 0, 0, 1), scale: taichi.math.vec3 = ti.math.vec3(1, 1, 1), color: taichi.math.vec3 = ti.math.vec3(1, 0, 1), opacity: taichi.f32 = 1) → rtgs.gaussian.Gaussian

Python scope Gaussian constructor to create a new Gaussian.

Parameters:

  • position – gaussian center (mean).

  • rotation – gaussian rotation quaternion.

  • scale – gaussian scale.

  • color – gaussian base color.

  • opacity – gaussian opacity.

Returns:

a Gaussian dataclass.