Doctoral Thesis: Efficient Generative Models for Visual Synthesis
32-G575 It will also be available on Zoom.
By: Tianwei Yin
Thesis Supervisor(s): Fredo Durand, William T. Freeman
Details
- Date: Thursday, March 13
- Time: 10:30 am - 12:30 pm
- Category: Thesis Defense
- Location: 32-G575 It will also be available on Zoom.
Additional Location Details:
While current visual generative models achieve remarkable quality, they struggle with high computational costs and latency, limiting their use in interactive applications. In this talk, I will present my research on improving the efficiency of generative models for image and video creation. I will begin by introducing distribution matching distillation, a technique that enables the training of one- or few-step visual generators by distilling knowledge from powerful yet computationally expensive diffusion models. Next, I will present improved distillation methods that enhance robustness and scalability, leading to a production-grade few-step image generator that is now deployed in widely used software, generating hundreds of millions of images annually. Finally, I will show how we can further reduce the latency for video generation, by switching to an autoregressive generation paradigm, enabling fast interactive video generation and world simulation.
Host
- Tianwei Yin
- Email: tianweiy@mit.edu