16-825 Assignment 1: Rendering Basics with PyTorch3D

1. Practicing with Cameras

360-degree render of the cow mesh
1.1 360-degree render.
Dolly zoom render of the cow on a plane
1.2 Dolly zoom.

2. Practicing with Meshes

360-degree render of a tetrahedron
2.1 Tetrahedron. The tetrahedron mesh has 4 vertices and 4 triangular faces.
360-degree render of a cube
2.2 Cube. The cube mesh has 8 vertices and 12 triangular faces.

3. Re-texturing a Mesh

I used color1 = [247/255, 37/255, 133/255] for the front of the cow and color2 = [76/255, 201/255, 240/255] for the back.

360-degree render of the retextured cow
Retextured cow.

4. Camera Transformations

Image R_relative T_relative Effect
cow_1 [[0, 1, 0], [-1, 0, 0], [0, 0, 1]] [0, 0, 0] Rotate around z-axis.
cow_2 I [0, 0, 2] Move farther away.
cow_3 I [0.5, -0.5, 0] Lateral shift.
cow_4 [[0, 0, 1], [0, 1, 0], [-1, 0, 0]] [-3, 0, 3] Side view.
Base camera transformation render
Base camera.
Camera transformation 1 render
Transform 1.
Camera transformation 2 render
Transform 2.
Camera transformation 3 render
Transform 3.
Camera transformation 4 render
Transform 4.

5. Rendering Generic 3D Representations

5.1 RGB-D Point Clouds

Point cloud from first RGB-D image
Point cloud from image 1.
Point cloud from second RGB-D image
Point cloud from image 2.
Combined point cloud from both RGB-D images
Union of both point clouds.

5.2 Parametric Functions

Parametric torus point cloud
Parametric torus point cloud.
Parametric Mobius strip point cloud
Additional parametric object: Mobius strip.

5.3 Implicit Surfaces

Meshes look smoother and more complete because they store surface connectivity, but extracting them requires voxelization and marching cubes. Point clouds are easier to sample and merge, but often need many points to avoid sparse or noisy renderings.

Implicit torus mesh
Implicit torus mesh.
Implicit heart mesh
Additional implicit object: heart mesh.

Extra Credit 6. Do Something Fun

Animated beating heart mesh
Beating implicit heart mesh.

Extra Credit 7. Sampling Points on Meshes

Original cow mesh Cow point cloud with 10 samples
Original mesh vs. 10 sampled points.
Original cow mesh Cow point cloud with 100 samples
Original mesh vs. 100 sampled points.
Original cow mesh Cow point cloud with 1000 samples
Original mesh vs. 1000 sampled points.
Original cow mesh Cow point cloud with 10000 samples
Original mesh vs. 10000 sampled points.