CMU 2022 Spring 16-889

Assignment 1: Rendering Basics with PyTorch3D

• Name: Chih-Wei Wu

• Email: chihweiw@andrew.cmu.edu

1. Practicing with Cameras

1.1. 360-degree Renders

1.2 Re-creating the Dolly Zoom

2. Practicing with Meshes

2.1 Constructing a Tetrahedron

• There are 4 vertices, 4 faces in the above mesh.

2.2 Constructing a Cube

• There are 8 vertices, 12 faces in the above mesh.

3. Re-texturing a mesh

• color1 = [1, 0, 1]
• color2 = [1, 1, 0]

4. Camera Transformations

• R_relative = [[0, 1, 0], [-1, 0, 0], [0, 0, 1]]
• T_relative = [0, 0, 0]

• R_relative = [[1, 0, 0], [0, 1, 0], [0, 0, 1]]
• T_relative = [0, 0, 3]

• R_relative = [[1, 0, 0], [0, 1, 0], [0, 0, 1]]
• T_relative = [0.5, -0.5, 0]

• R_relative = [[0, 0, 1], [0, 1, 0], [-1, 0, 0]]
• T_relative = [-3, 0, 3]

5. Rendering Generic 3D Representations

5.1 Rendering Point Clouds from RGB-D Images

5.2 Parametric Functions

5.3 Implicit Surfaces

• Tradeoff between mesh vs. point cloud:
  • Rendering speed: The rendering speed of point cloud is faster, while the rendering speed of mesh is slower.
  • Rendering quality: The rendering quality is better for mesh compared to point cloud, because the mesh uses faces to render the object surface while point cloud need vast amount of points to achieve the same surface smoothness.
  • Ease of use: This depends on the task. For example, if targeted for animation purpose, mesh is usally the go-to option because it is easier to animate an object by a few control points. If targeted for learning purpose, point cloud is usually easier to use than mesh.
  • Memory usage: Mesh consumes smaller memory to store compared to point cloud for achieving the same surface smoothness.

6. Do Something Fun

7. Sampling Points on Meshes

• Original
  

• Sample 10 points
  

• Sample 100 points
  

• Sample 1000 points
  

• Sample 10000 points