Q1 Practicing with Cameras

360-degree renders

Recreating the dolly zoom

Q2 Practicing with Meshes

Constructing a tetrahedron

Attribute	Count
Vertices	4
Faces	4

Constructing a cube

Attribute	Count
Vertices	8
Faces	12

Q3 Re-texturing a mesh

I chose color1 = [1, 0, 0] and color2 = [0, 0, 1] to invert the cow’s color from the example image provided.

Q4 Camera Transformations

Pure Rotation

Rotated by Z (pointing opposite to object, hence 90 clockwise)

R = [
    [0, -1, 0],
    [-1, 0, 0],
    [0, 0, 1]
]
T = [0, 0, 0]

Far

Moved by Z (pointing opposite to object, thus positive)

R = [
        [1, 0, 0],
        [0, 1, 0],
        [0, 0, 1]
]
T = [0, 0, 3]

Left

Moved by X (pointing right, thus positive)

R = [
        [1, 0, 0],
        [0, 1, 0],
        [0, 0, 1]
]
T = [0.5, 0, 0]

Side

Rotated by Y (pointing down, thus 90 Degrees), move by X and Z.

R = [
    [0, 0, 1],
    [0, 1, 0],
    [-1, 0, 0]
]
T = [-3, 0, 3]

Q5 Rendering Generic 3D Representations

Rendering Point Clouds from RGB-D Images

Parametric Functions

Number of Samples: 100

Number of Samples: 200

Number of Samples: 500

Implicit Surfaces

Voxel Size: 32

Voxel Size: 64

Voxel Size: 128

Tradeoffs between Parametric Functions and Implicit Surfaces

Parametric Function
Num Samples	100	200	*500*	1000
Time	11.29s	27.98s	*97.02s*	161.96s
Memory	595 MB	599 MB	*665 MB*	1099 MB
Implicit Surface
Voxel Size	32	64	128	256
Time	15.80s	*19.23s*	34.05s	111.23s
Memory	710 MB	*712 MB*	712 MB	774 MB

Quality of Parametric Torus at number of sampled points = 500 feels better to me than 100 or 200. Implicit Torus looks nice even at small voxel sizes, at 32 or 64. Thus, for a similar level of quality, the Implicit Surfaces are faster however, they consume more memory than Parametric functions.

Q6 Do Something Fun

Is it really a mesh or a point cloud? Take a closer look.

Assigment 1