Zeqi Gu

I am a third year PhD student at Cornell Tech, advised by Prof. Abe Davis and Prof. Noah Snavely. Before that I received my bachlor's degree from Cornell University with double major in computer science and mathematics.

My research interests lie in computer vision and graphics, with a focus in content creation using generative models. My past projects involve cartoon animation, style transfer, video decomposition and matting, and adversarial attacks, for which I've had the fortune to collaborate with Prof. Serge Belongie and Prof. Bharath Hariharan. I have interned at NVIDIA, Adobe, and Uber.

Email  /  CV  /  Github

profile photo
Filtered-Guided Diffusion: Fast Filter Guidance for Black-Box Diffusion Models
Zeqi Gu*, Ethan Yang*, Abe Davis
project page / arXiv

Fast, lightweight, and top-performing image-to-image translation method for diffusion models, motivated by signal processing (bilateral filtering).

FactorMatte: Redefining Video Matting for Re-Composition Tasks
Zeqi Gu, Wenqi Xian, Noah Snavely, Abe Davis
SIGGRAPH Journal, 2023
project page / arXiv

Extending video matting to scenes with complex foreground-background interactions.

Enhancing Adversarial Example Transferability with an Intermediate Level Attack
Qian Huang*, Isay Katsman*, Horace He*, Zeqi Gu*, Serge Belongie, Ser-Nam Lim
ICCV, 2019
project page / arXiv

An attack method that fine-tunes an existing adversarial example for greater black-box transferability by increasing its perturbation on a pre-specified layer of the source model.

Robotic Dough Shaping
Jan Ondras, Di Ni, Xi Deng, Zeqi Gu, Henry Zheng
ICCAS (Oral), 2022
project page / arXiv

A robot arm using vision and tacile information to roll a dough into a given shape. A course project turned into a paper.


      Paper reviewer for: AAAI (2022, 23), CVPR Workshop CV4ARVR (2022), and ICCV (2023).

      Teaching assistant for: CS 4670 (Spring 2022), CS 6670 (Fall 2021)


      In my spare time, I enjoy making videos about fashion history; I am also learning Spanish and tennis.

Thank Jon Barron for sharing his website's source code.