Top

‘It’s given me huge encouragement’: PhD grad wins 2022 Alain Fournier Dissertation Award for geometry processing research on stylization algorithms

Hsueh-Ti Derek Liu smiles facing the camera.

PhD graduate Hsueh-Ti Derek Liu is the recipient of the 2022 Alain Fournier Dissertation Award for his geometry processing research on stylization algorithms. (Photo: Supplied)

Hsueh-Ti Derek Liu has won the Alain Fournier Dissertation Award from the Canadian Human-Computer Communications Society (CHCCS).

The recent PhD graduate from the University of Toronto’s Department of Computer Science has been named the 2022 recipient of the prestigious national award for his thesis “Algorithms for Data-Driven Geometric Stylization & Acceleration” which explores the use of computer algorithms for creating 3D digital content and numerical tools for processing high-resolution geometric data.

The award is presented at the annual Graphics Interface conference for an outstanding doctoral dissertation completed at a Canadian university in the field of computer graphics. It is named in honour of the late Alain Fournier, a Canadian researcher who promoted excellence, both within Canada and internationally, in the field of computer graphics.

By taking a machine learning approach, Liu’s thesis demonstrates how algorithms and an optimization framework can serve as new modelling tools to lower the difficulties of stylizing 3D geometric objects.

“Let’s say, you scan yourself into some virtual world, and you have a 3D avatar of yourself and the stylization could mean for example, transforming your avatar into a cubic style, or Pixar style,” Liu explains.

The second part of his thesis looks at scalability — how the stylization process can be accelerated to run faster.

“We have these tools and we treat it as an interactive tool for users that you open in an app and then you can tweak your avatar. You don’t want to say, ‘I want to turn into a certain style’ and wait for an hour. So, this scalability part made sure we can run this algorithm on a very high-resolution model while maintaining efficient performance.”

Liu says the main application of this thesis is the metaverse — the concept of a single, shared, immersive, 3D virtual space — and points to the Disney movie Wreck-It Ralph in which characters can jump into virtual worlds and are transformed into a particular style.

“We hope that in the future when everything’s on the metaverse, wherever we have a 3D website and then we jump into a different website in a different style, we will immediately transform ourselves into the style of that world. How can we quantify the style of that world and after we know how to quantify that, how can we turn ourselves into that style, but still preserve our identity?” Liu explains. “The ultimate goal is to be in the metaverse where we can translate everything into different styles.”

The geometry processing researcher says he became interested in 3D shape manipulations and design early on and he particularly enjoys getting to see the end results of his work.

“We’re creating beautiful 3D shapes and renderings. So, it’s very satisfying to do this research, you can see all your results and it’s very pleasing. Sometimes I just fall in love with the results created by our algorithms,” he adds.

Hsueh-Ti Derek Liu’s thesis examined using algorithms for cubic stylization of 3D shapes in a virtual world while maintaining textures and geometric features. (Credit: Hsueh-Ti Derek Liu, 3D assets obtained from sketchfab.com by smeerws and Jesús Orgaz licensed under CC BY 4.0.) 

Liu was supervised by Associate Professor Alec Jacobson.

“Derek is an exceptional graduate of our PhD program, demonstrating excellence in research, a dedication to outreach and a caring approach to mentoring junior students. His thesis is a valuable contribution to the field of computer graphics and 3D learning, serving as a reference and an inspiring example for future students. As Derek’s PhD supervisor, it was an honour to witness his growth and accomplishments,” says Jacobson.

While at U of T, Liu along with other graduate students and advised by Jacobson, started the Toronto Geometry Colloquium as a platform to help promote young and underrepresented researchers.

Through a weekly web series format, these researchers were given a stage to practice their presentation skills and showcase their work in geometry processing.

He says he learned while completing his PhD that planning is key and encourages students not to rush their ideas.

“Something I found very useful is to plan ahead when you work on something. Let’s say the deadline is January, in September, you should start to think about what you want to do next, even when you are still working on the same project. I think it’s very important to work on ideas that you polish for a long time, a couple of months, or even a year or more,” Liu says.

He attributes his positive experience at U of T to his advisor and the lab environment he worked in, which gave him the opportunity to problem solve and dive into a topic he loves.

“It changed me a lot and I really appreciate having this period of time.”

Liu says when he started his PhD, he never imagined he would receive this award, especially when he questioned at times whether the work he was doing was important to the community.

“It’s really an honour to receive this,” he reflects. “It’s given me huge encouragement to continue to research into this topic and hope one day in the future we can see Wreck-It Ralph stylization become something that’s happening in practice.”

In his role as a research scientist at Roblox, he continues to conduct computer graphics research with the goal of making 3D content creation/stylization accessible to everyone. This objective feeds into the company’s ambitions for everyone to build immersive 3D experiences in the metaverse.

He says he is interested in working directly on 3D-to-3D stylization, which he believes is essential for the metaverse.

“I hope we can build that stylization tool or a shape manipulation tool so that a user without professional experience can still make content they really want in a certain style,” he explains. “I hope my skills can contribute to making the metaverse happen.”