16-825 Assignment 1: Rendering Basics with PyTorch3D

Sarah Pendhari Andrew ID: spendhar

1. Practicing with Cameras

1.1 360-Degree Renders

360 degree turntable of the cow mesh
Cow mesh, 360° turntable

1.2 Re-creating the Dolly Zoom

Dolly zoom effect on the cow mesh
Dolly zoom

2. Practicing with Meshes

2.1 Constructing a Tetrahedron

360 degree turntable of a tetrahedron mesh
Tetrahedron, 360° turntable

Vertices: 4    Faces: 4

2.2 Constructing a Cube

360 degree turntable of a cube mesh
Cube, 360° turntable

Vertices: 8    Faces: 12

3. Re-texturing a Mesh

color1 = [0, 0, 1] (blue), color2 = [1, 0, 0] (red).
Cow mesh retextured with a front-to-back color gradient
Front-to-back color gradient retexture

4. Camera Transformations

Base render (R_0, T_0):

Base cow render with R_0, T_0
Base view
Camera transformation 1
Transformation 1
R_relative rotates the camera by 90 degrees about the Z-axis while T_relative is zero. Therefore, the camera position remains unchanged while its orientation is rotated.
Camera transformation 2
Transformation 2
R_relative is the identity matrix, so the camera orientation remains unchanged. T_relative = [0, 0, 2] translates the camera farther along the Z direction, increasing its distance from the cow and making the cow appear smaller.
Camera transformation 3
Transformation 3
R_relative is the identity matrix, so there is no camera rotation. T_relative = [0.5, -0.5, 0] translates the camera in the X and Y directions, changing the position of the cow in the image while preserving its orientation.
Camera transformation 4
Transformation 4
R_relative rotates the camera about the Y-axis so that it views the cow from its side. T_relative = [-3, 0, 3] compensates for the change in camera position caused by the rotation and places the camera on the side of the cow while keeping the cow in view.

5. Rendering Generic 3D Representations

5.1 Rendering Point Clouds from RGB-D Images

Side by side point clouds: image 1, image 2, and their union
Left to right: point cloud from image 1, image 2, and their union

5.2 Parametric Functions

360 degree turntable of a torus point cloud
Torus point cloud (parametric)
360 degree turntable of a Mobius strip point cloud
Bonus: Möbius strip point cloud

5.3 Implicit Surfaces

360 degree turntable of a torus mesh from an implicit function
Torus mesh (implicit, marching cubes)
360 degree turntable of a metaball mesh
Bonus: metaball mesh

Mesh vs. point cloud tradeoffs

6. Extra Credit: Do Something Fun

A cat!

360 degree turntable of a cat mesh
Cat, 360° turntable (from Free3D)

7. Extra Credit: Sampling Points on Meshes

Stratified sampling of the cow mesh's surface: sample a face with probability proportional to its area, sample a uniform random barycentric coordinate, then compute the corresponding point. Shown below (left to right): the original mesh, then point clouds sampled with 10, 100, 1000, and 10000 points.

Cow mesh next to point clouds sampled at 10, 100, 1000, and 10000 points
Original mesh, then 10 / 100 / 1000 / 10000 sampled points