16-825 Assignment 1: Rendering Basics with PyTorch3D (Total: 100 Points + 10 Bonus)

This page includes text, an image, and GIFs using my own media.

Task 1.1 — 360

Result:

Task 1.2 — 3D Shape Image

Result:

Task 2.1 — tetrahedron

Result:

Task 2.2 — cube

Result:

Task 3 — retexture the cow!

Result:

cotton candy cow — colorful cow colored magenta and cyan because of CMYK without YK

Task 4 — transform the cow!

transform1:
R_relative = [[0, 1, 0], [-1, 0, 0], [0, 0, 1]]
T_relative = [0, 0, 0]

transform2:
R_relative = [ [0.0, 0.0, 1.0], [0.0, 1.0, 0.0], [-1.0, 0.0, 0.0]]
T_relative = [-3, 0, 3]

transform3:
R_relative = [[1, 0, 0], [0, 1, 0], [0, 0, 1]]
T_relative = [0, 0, 3]

transform4:
R_relative = [[1, 0, 0], [0, 1, 0], [0, 0, 1]]
T_relative = [0.5, -0.5, -0.2]

R_relative and T_relative should be rotating and translating the camera, with the rotation applied first and the translation applied wrt to the camera post rotation.

Task 5.1 — plants plants plants

Result:

The point cloud corresponding to the first image.

The point cloud corresponding to the second image.

The point cloud formed by the union of the first 2 point clouds.

TThe point clouds side by side — The point cloud from above side by side.

Task 5.2 — parametric

Result:

a trianguloid math from https://paulbourke.net/geometry/tranguloid/

Task 5.3 — implicit

Result:

a nose — A nose. from https://paulbourke.net/geometry/nose/

Tradeoffs:

Rendering as a mesh is faster but certain details like reflectance/sphecular hightlights might be lost.
However, generating proper clean manifold meshes can be quite difficult as connectivity can be hard to keep track of.

Rendering as a point cloud might be able to get more details, with certain shapes easier to compute than its mesh counterpart,
you need a lot of points to create a similar shape to a mesh.

Task 6 — funfunfun

Result: