Mapping the space of motion diffusion models to optimize performance

This research project focuses on optimizing the efficiency and performance of motion diffusion models for real-time applications in video games. Diffusion models have shown great potential in producing high-quality and diverse human motion animations, but are often limited by their computational demands. This project will explore different model architectures and optimization techniques to reduce their resource consumption, making them faster and more memory-efficient without sacrificing animation quality. The project will aid CD Projekt Red in deploying optimized motion diffusion models for games in production, affirming its position as a technical innovator in the gaming industry while providing material benefits through the development of an evaluation pipeline for future model assessments.

Faculty Supervisor:

KangKang Yin

Student:

Partner:

CDPR Canada

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

Simon Fraser University

Program: