Course information

Time and location: TR 11:10 - 12:25 in 126 HRBB
Office hours: T 10:00 - 11:00
Office location: 527B HRBB
Email: nimak@tamu.edu

Overview

In this course, we first cover the basic principles of deep learning, including different network architectures and training strategies. We then study the recent use of deep learning in a variety of computer graphics applications such as view synthesis, image inpainting, colorization, and Monte Carlo denoising. Specifically, we discuss different solutions for this challenging problem, including novel system designs and dataset preparation strategies.

Prerequisites

There are no prerequisites. However, basic understanding of machine learning and computer graphics is recommended. Basic knowledge of Python is also recommended, but not necessary.

Course format

This is a graduate level course and it mainly consists of students presenting recent research papers on the applications of deep learning in computer graphics. For the first few classes, the instructor covers the basics of deep learning. Afterwards, each class consists of two presentations by students. Each presentation should be 20 minutes with an extra 15 minutes for class discussion.

We will have two assignments, which need to be implemented in one of the deep learning packages (PyTorch, TensorFlow, etc.). We will also have a final project, which could be anything related to the use of deep learning in computer graphics.

Textbook

There is no textbook for this course, since it is based on recent research papers. The papers are all available online.

Grading

Presentations (20%): Students are expected to present recent research papers during the semester.
Attendance and participation (10%): This is a graduate level course and, thus, students are expected to attend the class and participate in the discussions. Absences will be handled according to student rule 7. For more information click here.
Programming assignments (30%): There will be 2 programming assignments for this course.
Final project (40%): Students are free to choose any project that addresses a computer graphics application using deep learning. They can work on the project individually or in teams. Students will propose the project at the middle of the semester. The final project presentation as well as a report will be due at the end of the semester.

Schedule*

Date	Topic	Papers	Presenter(s)	Assignments
Aug 28	Introduction and overview		Nima
Aug 30	Basics of deep learning I		Nima
Sep 4	Basics of deep learning II		Nima
Sep 6	Style transfer	Gatys 16, Ruder 16	Tushar, Duc
Sep 11	Image synthesis I	Radford 15, Isola 17	Prateek, Nitin
Sep 13	Image synthesis II	Chen 17, Qi 18	Saurabh, Qinbo
Sep 18	View Synthesis I	Zhou 16, Flynn 16	Nishant, Avinash	Hw 1 out
Sep 20	View Synthesis II	Kalantari 16, Wu 17	Avinash, Yaohua
Sep 25	Colorization I	Larsson 16, Lizuka 17	Buvaneish, Buvaniesh
Sep 27	Colorization II	Zhang 17, He 18	Stuti, Peng
Oct 2	High dynamic range imaging I	Kalantari 17, Wu 18	Duc, Yaohua
Oct 4	High dynamic range imaging II	Eilertsen 17, Endo 17	Yerania, Saurabh	Hw 1 due
Oct 9	Image inpainting	Pathak 16, Lizuka 17	Yerania, Yashwanth	Hw 2 out
Oct 11	Image super-resolution	Dong 14, Ledig 17	Prateek, Samia
Oct 16	Sketches I	Wang 15, Sangkloy 17	Stuti, Fieza
Oct 18	Sketches II	Sangkloy 17, Simo-Serra 16	Fieza, Peng
Oct 23	Final project proposals I
Oct 25	Final project proposals I			Hw 2 due
Oct 30	Intrinsic decomposition I	Narihira 15, Shi 17	Cassandra, Cassandra
Nov 1	Intrinsic decomposition II	Liu 17, Bi 18	Yashwanth, Peeyush
Nov 6	Monte Carlo denoising I	Kalantari 15, Bako 17	Logan, Brennen
Nov 8	Monte Carlo denoising II	Chaitanya 17, Vogels 18	Nitin, Brennen
Nov 13	Shading	Nalbach 17, Thomas 17	Kevin, Logan
Nov 15	Relighting	Ren 13, Ren 15	Xiaojing, Tushar
Nov 20	Image denoising	Gharbi 16, Mildenhall 18	Xiaojing, Peeyush
Nov 22	Thanksgiving - no class
Nov 27	Other topics I	Zhou 18, Xu 18	Kevin, Samia
Nov 29	Other topics II	Ephrat 18, Prashnani 18	Qinbo, Nishant
Dec 4	Final project presentation I
Dec 6	Final project presentation II

*Schedule is subject to small changes.