[16-889] Learning for 3D Vision, Spring 2022

Assignment 1

Excuse the visual artifacts in the gifs due to compression!

1. Practicing with Cameras

1.1. 360-degree Renders (5 points)

x
python main.py --qn 1_1 --mesh_path <path/to/mesh> --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...>

1.2 Re-creating the Dolly Zoom (10 points)

x
python main.py --qn 1_2 --mesh_path <path/to/mesh> --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...>

2. Practicing with Meshes

2.1 Constructing a Tetrahedron (5 points)

Constructing a tetrahedron requires 4 faces and vertices each.

x
python main.py --qn 2_1 --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...>

2.2 Constructing a Cube (5 points)

Constructing a cube requires 12 faces and 8 vertices.

python main.py --qn 2_2 --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...>

3. Re-texturing a mesh (10 points)

The colors used were: color1 = [0.0, 1.0, 0.0] and color2 = [1.0, 0.0, 1.0].

x
python main.py --qn 3 --mesh_path <path/to/mesh> --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...>

4. Camera Transformations (20 points)

R_relative	$\left[\begin{array}{ccc}0& 1& 0\\ -1& 0& 0\\ 0& 0& 1\end{array}\right]$$\begin{bmatrix} 0 & 1 & 0\\ -1 & 0 & 0\\0 & 0 & 1\end{bmatrix}$	$\left[\begin{array}{ccc}0& 0& 1\\ 0& 1& 0\\ -1& 0& 0\end{array}\right]$$\begin{bmatrix} 0 & 0 & 1\\ 0 & 1 & 0\\-1 & 0 & 0\end{bmatrix}$	$\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right]$$\begin{bmatrix} 1 & 0 & 0\\ 0 & 1 & 0\\0 & 0 & 1\end{bmatrix}$	$\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right]$$\begin{bmatrix} 1 & 0 & 0\\ 0 & 1 & 0\\0 & 0 & 1\end{bmatrix}$
T_relative	$\left[\begin{array}{ccc}0& 0& 0\end{array}\right]$$\begin{bmatrix} 0 & 0 & 0\end{bmatrix}$	$\left[\begin{array}{ccc}-3& 0& 3\end{array}\right]$$\begin{bmatrix} -3 & 0 & 3\end{bmatrix}$	$\left[\begin{array}{ccc}0& 0& 3\end{array}\right]$$\begin{bmatrix} 0 & 0 & 3\end{bmatrix}$	$\left[\begin{array}{ccc}0.5& -0.5& 0\end{array}\right]$$\begin{bmatrix} 0.5 & -0.5 & 0\end{bmatrix}$

T_relative and R_relative describes an affine transformation (only R, T) for the camera with respect to the world origin.

5. Rendering Generic 3D Representations

5.1 Rendering Point Clouds from RGB-D Images (10 points)

xxxxxxxxxx
python main.py --qn 5_1 --image_size <256/512/...> --fps <10/24/...>

5.2 Parametric Functions (10 points)

x
python main.py --qn 5_2 --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...> --num_samples <100/1000/...>

5.3 Implicit Surfaces (15 points)

xxxxxxxxxx
python main.py --qn 5_3 --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...>

Rendering a mesh is easier because pointclouds need to be converted to a 3D primitive (sphere, cube) first before they can be rasterized thus incurring an additional compute + memory overhead. Since meshes come with face information, the shading and blending step calculates the interpolated color/feature values for each face; while this step has to be done for each 3D primitive corresponding to each point. This implies that rendering large meshes would be faster than rendering large pointclouds (with same no. of vertices).

6. Do Something Fun (10 points)

6. 1 Adding material properties

One can add material properties to change the manipulate rendering of a mesh. Increasing the shininess component allows for specular highlights on the mesh.

x
python main.py --qn 6_1 --mesh_path <path/to/mesh> --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...>

6. 2 Render surface normal maps

To render surface normal maps of meshes, one has to write a custom shader leaving the rasterization step as it is. The shader retrieves the face corresponding to each pixel (assuming no blending) and calculates the normal of each face (by averaging the corresponding vertex normals).

python main.py --qn 6_2 --mesh_path <path/to/mesh> --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...>

(Extra Credit) 7. Sampling Points on Meshes (10 points)

x
python main.py --qn 7 --mesh_path <path/to/mesh> --image_size <256/512/...> --output_file <path/to/file> --fps <10/24/...> --num_samples <100/1000/...>