Render the cow from all angles to make a spinning gif.
Change the field of view while moving the camera so the cow stays the same size.
Build a tetrahedron by hand (4 vertices, 4 faces) and spin it.
Build a cube by hand (8 vertices, 12 triangle faces) and spin it.
Color the cow from front to back, blending blue at the front into red at the back.
Find the camera moves that reproduce each target view.
Rotate in the image plane
R_relative = [ 0 1 0 ] T_relative = [0, 0, 0]
[-1 0 0 ]
[ 0 0 1 ]
View from the side
R_relative = [ 0 0 1 ] T_relative = [-3, 0, 3]
[ 0 1 0 ]
[-1 0 0 ]
Move the camera back so the cow looks smaller
R_relative = [ 1 0 0 ] T_relative = [0, 0, 3]
[ 0 1 0 ]
[ 0 0 1 ]
Shift the cow left and down
R_relative = [ 1 0 0 ] T_relative = [0.4, -0.4, 0]
[ 0 1 0 ]
[ 0 0 1 ]
Turn two color plus depth images into 3D point clouds and view each one plus their union.
Sample a torus and a hyperboloid from their parametric equations as point clouds.
Turn implicit equations into meshes (marching cubes): a torus (hole visible) and a hollow cube frame.
Mesh vs. point cloud: I look at the tradeoff to be something between connectivity and smoothness. A mesh size definitely depends on the size we choose, it is often coarse and takes a long time to render if a sharper resolution is chosen, it has the connectivity data. If someone is prioritizing for connectivity data, mesh would make more sense. Point clouds are able to give an idea of the shape and smoothness quicker based on the sampling, as the points don't need to store connectivity, the surface and connectivity data might not be there, at the same time for just looking at shape or representation, this would be much better.
Paint a grid of random colors onto the cow through its UV texture.
Uniformly sample the cow's surface and compare 10 / 100 / 1000 / 10000 points to the mesh.