⎡ 0 1 0 ⎤ ⎡ 0 ⎤
R_relative = ⎢-1 0 0 ⎥ T_relative = ⎢ 0 ⎥
⎣ 0 0 1 ⎦ ⎣ 0 ⎦
R_relative rotates the camera 90 degrees around the z-axis while T_relative leaves its position unchanged.
⎡ 1 0 0 ⎤ ⎡ 0 ⎤
R_relative = ⎢ 0 1 0 ⎥ T_relative = ⎢ 0 ⎥
⎣ 0 0 1 ⎦ ⎣ 1.5 ⎦
R_relative keeps the camera orientation unchanged while T_relative moves the camera from world z = -3 to -4.5.
⎡ 1 0 0 ⎤ ⎡ 0.5 ⎤
R_relative = ⎢ 0 1 0 ⎥ T_relative = ⎢-0.5 ⎥
⎣ 0 0 1 ⎦ ⎣ 0 ⎦
R_relative keeps the camera orientation unchanged while T_relative shifts the view by 0.5 in x and -0.5 in y.
⎡ 0 0 1 ⎤ ⎡-3 ⎤
R_relative = ⎢ 0 1 0 ⎥ T_relative = ⎢ 0 ⎥
⎣-1 0 0 ⎦ ⎣ 3 ⎦
R_relative rotates the camera 90 degrees around the y-axis while T_relative repositions it to keep the cow in view.


A torus and a spring


Meshes usually give better visual quality than sparse point clouds, with continuous surfaces and smooth lighting. Point clouds are simpler; speed and memory depend on triangle/point counts, while implicit extraction requires a voxel grid.
Cow being orbited by 2 rings.
Left to right: 10, 100, 1,000, 10,000 points, and the original mesh.