16-889: Learning for 3D Vision

Assignment 1

Adnan Ahmad (adnana)

Late days used - 1

1. Practicing with cameras

1.1

1.2

2 Practicing with Meshes

2.1

Number of Vertices : 4

Number of Faces : 4

2.2

Number of vertices: 8

Number of faces : 12

3. Retexturing a mesh

4. Camera Transformations

Here we have to rotate the camera axis such that the given y axis becomes -x-axis (minus) and x-axis to be shifted to given y-axis location. There is not translation

R = [ [ 0, 1, 0],

         [-1, 0, 0],

         [ 0 0, 1]]

T = [ 0, 0, 0]

Here we have to translate the camera axis such that the given camera center now lies in positive new Z-axis. There is no rotation.

R = [ [ 1, 0, 0],

         [ 0, 1, 0],

         [ 0 0, 1]]

T = [ 0, 0, 2]

Here we have to translate the camera axis such that the given camera center now lies in positive new X and negative new Y. There is no rotation.

R = [ [ 1, 0, 0],

         [ 0, 1, 0],

         [ 0 0, 1]]

T = [ 0.5, -0.5, 0]

Here we have to rotate the camera axis such that the new X axis becomes given - Z-axis(minus) and new Z-axis to be shifted to given X-axis location. Here we have to translate the camera such that the given camera center now lies in positive new Z and positive new X.

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

         [ 0, 1, 0],

         [ 1 0, 0]]

T = [ -3, 0, 3]

5. Rendering Generic 3D

5.2 Parametric Functions

5.3 Implicit Function

Discussing Tradeoff between Point Cloud and Meshes

Rendering Speed:

For most complex structures, meshes are faster to render than point clouds since we can represent them in fewer number of datapoints when compared to dense point clouds

Rendering Quality:

The rendering on meshes tends to better since they form closed structures and don’t leave open spaces which may be a problem in non densely sampled point clouds

Ease of Use:

Meshes while intuitive for easy shapes can be difficult to use for complex shapes and one can miss connectivity of vertices, while Point clouds consisting of only 3D points of objects are a bit easy to use.

Memory Usage:

This depends on the resolution of our input representation, if we want really detailed representations then a very dense point cloud will take more memory than an equivalent mesh. However for a course input, the number of points may be less and in those cases a point cloud may be more memory efficient.

6 Do Something Fun

I Visualised a 3D model of my favourite car.