Learning for 3D Vision: Assignment 1

1. Practicing with Cameras

1.1. 360 Degree Rendering

1.2 Re-creating the Dolly Zoom

2. Practicing with Meshes

2.1. Constructing a Tetrahedron

2.2. Constructing a Cube

3. Re-texturing a mesh

color1 is (0.,0.6,0.4) and color2 is (1.0,0.0,0.0)

4. Camera Transformations

R_rel and T_rel rotates and translates the camera in the new orientation w.r.t the world coordinate. The corresponding values for each of the below rendering are written in the code.

5. Rendering Generic 3D Representations

5.1 Rendering Point Clouds from RGB-D Images

First image	Second image	Union of the first 2 point clouds.

5.2 Parametric Functions

5.3 Implicit Surfaces

Point clouds are fast to render compared to mesh as the points can be simply visualized in the 3D space instead of rendering the surfaces as well. Rendering quality will be better with mesh with similar number of vertices/point cloud size. For the same rendering quality more points are required in a point cloud. Point cloud is easier to use when as an input to neural networks and defining distance functions between two point clouds. For the same number of points mesh need to store extra information regarding faces, thus are more memory intensive.

6. Do Something Fun

Parametrics functions can be used to create animations for smoothly changing shapes. Here I change a radial torus to elliptic torus continuously by changing the radius in one of the axis. The texture map is based on the x value of the parametric function.

7. Sampling Points on Meshes

number of late days used