Hello, I'm Thuan Hoang Nguyen

I am a PhD student at MBZUAI, advised by Prof. Hao Li.

Previously, I was a research resident at VinAI Research, where I worked under supervision of Dr. Anh Tran. Before that, I recieved a B.Eng. degree in Computer Science at the Ho Chi Minh University of Technology - VNUHCM where I was advised by Prof. Duc Dung Nguyen.


News

  • Aug 2024: I have joined MBZUAI as a PhD student.
  • Jul 2024: SwiftBrush v2 has been accepted to ECCV 2024.
  • Jun 2024: We have released the code for SwiftBrush.
  • Mar 2024: SwiftBrush has been accepted to CVPR 2024.
  • Jul 2023: Anti-DreamBooth has been accepted to ICCV 2023. We have also released the code.
  • Mar 2023: CREPS has been accepted to CVPR 2023. We have also released the code.
  • Jul 2022: I have joined VinAI Research as a research resident.

Publications
* means equal contribution

SwiftBrush v2: Make Your One-step Diffusion Model Better Than Its Teacher

SwiftBrush v2: Make Your One-step Diffusion Model Better Than Its Teacher

ECCV, 2024

An improved SwiftBrush that makes the one-step diffusion student beat its multi-step teacher.

SwiftBrush: One-Step Text-to-Image Diffusion Model with Variational Score Distillation

SwiftBrush: One-Step Text-to-Image Diffusion Model with Variational Score Distillation

Thuan Hoang Nguyen, Anh Tran
CVPR, 2024

An image-free distillation method that transforms multi-step text-to-image diffusion models into one-step generators.

Anti-DreamBooth: Protecting Users from Personalized Text-to-Image Synthesis

Anti-DreamBooth: Protecting Users from Personalized Text-to-Image Synthesis

ICCV, 2023

A security booth safeguards your privacy against malicious threats of DreamBooth by preventing it from synthesizing photo-realistic images of you.

Efficient Scale-Invariant Generator with Column-Row Entangled Pixel Synthesis

Efficient Scale-Invariant Generator with Column-Row Entangled Pixel Synthesis

Thuan Hoang Nguyen*, Thanh Van Le*, Anh Tran
CVPR, 2023

An efficient and geometry-invariant generator for synthesizing high-resolution images without using spatial convolutions or a coarse-to-fine design.