CSC 2547 Current Topics in Machine Learning Methods in 3D and Geometric Deep Learning

Instructor: Animesh Garg
Webpage: https://pairlab.github.io/csc2547-w21/
TAs: Dylan Turpin and Jun Gao

Winter 2021 Logistics

Lectures: Tues 5-7pm ET, Zoom
Office Hours: AG: Mon 3-4pm, Zoom
Discussion: Piazza (in Quercus)
Course Contact Email: garg@cs.toronto.edu
Email Subject: “CSC2547-w21:”

Description

This course introduces deep learning methods and modern advances in 3D Vision. We will study representations, learning algorithms and generative models for 3D vision tasks at object and scene level. We will then study Geometric Deep Learning and concepts of Manifold Learning as relevant to Deep Learning. The 3D nature of this topic has many potential applications in graphics, robotics, content creation, mixed reality, biometrics, and more.

Learning objectives

At the end of this course, you will be familiar with: 1. Representation, Learning and Generative Deep Learning of Point Cloud Input 2. Differentiable and Neural Rendering 3. Geometric Deep Learning 4. Applications of Non-Euclidean Deep Learning outside Computer Vision.

Textbook & Resources

Grading & Evaluation

This course will consist of lectures, along with paper presentations & discussions. Along with this there would be take home midterms, and a group project.

In-Class Paper Presentation: 25%
Paper Summary video (after feedback from in-class presentation): 10%
Take-Home Midterm: 15%
Project: 50% (5% proposal, 5% mid-term report, 15% final presentation, 25% final report)

Prerequisites

You need to be comfortable with: introductory machine learning concepts (such as from CSC411/ECE521 or equivalent), linear algebra, basic multivariable calculus, intro to probability. You also need to have strong programming skills in Python.

Optional, but recommended: experience with neural networks, such as from CSC 421/2516, introductory-level familiarity with computer vision.

Note: if you don't meet all the prerequisites above please contact the instructor by email.