Dave Epstein.
Toggle dark mode. Scholar. CV. Pixels. dave [at] eecs.berkeley.edu.
I am a final-year PhD student at Berkeley AI Research advised by Alexei Efros. My research is in unsupervised deep learning, with a focus on scalable, algorithmically interpretable generative architectures and objectives. I'm currently working on emergent structure and control in large models.
In the past, I have been fortunate to collaborate with Taesung Park, Richard Zhang, and Eli Shechtman at Adobe, as well as with Ben Poole, Aleksander Holynski, Chen Sun, Jiajun Wu, and Cordelia Schmid at Google. I graduated in 2020 from Columbia with a BS in computer science, where I was lucky to be introduced to machine learning and advised by Carl Vondrick. My research has been supported by the PD Soros Fellowship.
Research.
-
Disentangled 3D Scene Generation with Layout Learning.
Dave Epstein, Ben Poole, Ben Mildenhall, Alexei A. Efros, Aleksander Holynski.
PaperProject page -
Diffusion Self-Guidance for Controllable Image Generation.
Dave Epstein, Allan Jabri, Ben Poole, Alexei A. Efros, Aleksander Holynski.
PaperProject page Talk Demo
NeurIPS 2023. -
BlobGAN: Spatially Disentangled Scene Representations. NEW!
Dave Epstein, Taesung Park, Richard Zhang, Eli Shechtman, Alexei A. Efros.
PaperProject page Talk Code Demo
ECCV 2022. -
Globetrotter: Unsupervised Multilingual Translation from Visual Alignment. NEW!
Didac Suris, Dave Epstein, Carl Vondrick.
PaperProject page
CVPR 2022 (Oral). -
Learning Temporal Dynamics from Cycles in Narrated Video.
Dave Epstein, Jiajun Wu, Cordelia Schmid, Chen Sun.
PaperProject Page Talk Blogpost
ICCV 2021. -
Learning to Learn Words from Visual Scenes.
Didac Suris*, Dave Epstein*, Heng Ji, Shih-Fu Chang, Carl Vondrick
PaperProject pageTalk
ECCV 2020. -
Oops! Predicting Unintentional Action in Video.
Dave Epstein, Boyuan Chen, Carl Vondrick.
PaperProject PageTalk
CVPR 2020.
Teaching.
At Columbia.
Head Teaching Assistant.
Head Teaching Assistant.