Edy Garfinkiel
Edy Garfinkiel
Dynamic Graphics Project
Department of Computer Science
University of Toronto

e-mail: {first letter of first name, first seven letters of last name} at cdf dot toronto dot edu

40 St. George Street
Toronto, ON
M5S 2E4

About Me:
I'm currently the instructor for CSC418 (Computer Graphics), UofT, course page here.
Previously I was a Software Engineer at LinkedIn.
Previously, I was a master's student in the Dynamic Graphics Project (DGP) lab at the University of Toronto in the Department of Computer Science.


My Resume (available upon request)
University of Toronto
M.Sc. Computer Science (Computer Graphics), 2013, University of Toronto.
B.Math. Honours, Computer Mathematics Information Technology (Minor in Economics), 2010, Carleton University.
Master's Research:
I worked under the supervision of Aaron Hertzmann and Pierre Bénard. My research in non-photorealistic rendering (NPR) centered on style transfer in line drawings. The theory is that style should be separable from its underlying base path. Once the style is decoupled, it should be possible to have it reapplied to another underlying base bath. The motivation for this work comes from line drawings where different artists draw the same characters with different styles. On the left hand side is an example of a line drawing of a bear that is drawn in two different styles. By decoupling the style, an artist's style can be captured and applied to the lines drawn by a novice user, as an example. The approach taken in this work is to use curvature as the main feature of style.

Technical report

Course Projects:

Introduction to Machine Learning (CSC2515):

For the final project, I was motivated to find a way to learn something interesting about styles of strokes. Since I did not have easily available data for line drawing strokes, I reverted to using characters, with the style being their font. The characters in each font were converted to an image large enough to capture enough detail of the style. I chose a set of fonts that had a very stylistic design and trained three neural network architectures to learn a mapping from the original image space of the character to a new "font" space. The idea in mind was to see if the characters of the same font would cluster in this new space, with a very small amount of data (62 characters per font). Surprisingly, when visualized in two dimensions, even the simplest architecture showed obvious signs of clustering. Another motivation for the project was to try to find similar fonts. That is, if a network was given a font it had never seen then it would try to classify it as the "closest" available font.

Final report
Computer Graphics (CSC418/2504):

As part of a team of two, created an interactive world. This project was free-form and consisted of a large textured ground along with various obstacles placed on it. On the ground, several jointed, moving targets were placed with bullseye images. The world was wrapped around with a skybox. The user could navigate through this world via mouse and keyboard input and shoot the targets. Crosshairs were placed in the middle of the screen aiding in the aiming of the arrow. By clicking and holding, the cross hair would decrease in size and the initial speed of the arrow would increase. Among the winners of the Wooden Monkey Prize!

The Business of Software (CSC454/2527):

In a group of five graduate students over the course of a semester, worked together to form the completion of a full business plan for a software startup. The motivation behind the company was to provide better healthcare for patients by providing doctors and nurses access to a web-based tablet application to aid them in the decision-making process. Everything from product description, marketing, competition research, professional interviews to a financial model was worked on for the final document. My personal contributions were in the financial model, financial roadmap along with the final business model 'canvas'.

Final document
Honours Project, Carleton University School of Mathematics and Statistics:
Supervisor: Dr. Brian Mortimer

For my honours project I chose to focus on the topic of geometry, particularly on ellipses. I studied some of their reflective properties and was especially interested in determining whether an observation that I had made was in fact true. Namely, that multiple reflections through the foci of an ellipse eventually converge to a a path along the major axis. Through lots of reading on the area of Mathematical Billiards, I found some applicable works which allowed me to prove the conjecture.

Real-time 3D Game Engines (COMP 4002):

The project for this course consisted of implementing various parts of a game engine written fully in C++ with OpenGL. Firstly, we had to design a world for a level in a game in a map creation program. Rotator and translator objects were implemented along with caustics for a pool, sprites and a skybox. In the last part of the project, a terrain with a height map was implemented. Some user controls were as well added such as a help menu. This project gave me experience with programming in the large and reading and understanding the workings of code begun by another developer.

Some more information about me:
  • I am multilingual. I speak fluent English, Spanish and advanced French.
  • I do improv regularly. It's a great way to keep me fresh.