Trung Dao

Greetings! I am a Research Resident at VinAI Research, where I am fortunate to work under the mentorship of Dr. Anh Tran and Dr. Cuong Pham. My primary research interests focus on Deep Generative Models, including GANs and Diffusion Models. Previously, I worked as an AI Engineer also at VinAI, gaining hands-on experience in training, deploying, and optimizing deep learning models. During this period, I contributed to projects such as Face Recognition, Traffic Sign and Light Recognition, and Noise Cancelling, mostly deployed on edge devices, bridging the gap between research and real-world applications.

🌟 News

[12/2024] Self-Corrected Flow has been accepted to AAAI 2025.
[09/2024] DiMSUM has been accepted to NeurIPS 2024.
[07/2024] SwiftBrushV2 has been accepted to ECCV 2024.
[03/2024] EFHQ has been accepted to CVPR 2024.

Publications
* means equal contribution

Self-Corrected Flow Distillation for Consistent One-Step and Few-Step Image Generation

Self-Corrected Flow Distillation for Consistent One-Step and Few-Step Image Generation

Authors: Quan Dao*, Hao Phung*, Trung Dao, Dimitris Metaxas, Anh Tran

AAAI, 2025

A comprehensive distillation framework for latent flow matching models that excels in generating high-quality and consistent images in both one-step and few-step sampling

arXiv
DiMSUM : Diffusion Mamba - A Scalable and Unified Spatial-Frequency Method for Image Generation

DiMSUM : Diffusion Mamba - A Scalable and Unified Spatial-Frequency Method for Image Generation

Authors: Hao Phung*, Quan Dao*, Trung Dao, Hoang Phan, Dimitris Metaxas, Anh Tran

NeurIPS, 2024

A hybrid Mamba-Transformer diffusion model that synergistically leverages both spatial and frequency information for high-quality image synthesis.

Project Page arXiv GitHub
SwiftBrush v2: Make Your One-step Diffusion Model Better Than Its Teacher

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

Authors: Trung Dao, Thuan Hoang Nguyen*, Thanh Van Le*, Duc Vu*, Khoi Nguyen, Cuong Pham, Anh Tran

ECCV, 2024

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

Project Page arXiv GitHub
EFHQ: Multi-purpose ExtremePose-Face-HQ dataset.

EFHQ: Multi-purpose ExtremePose-Face-HQ dataset.

Authors: Trung Dao*, Duc Vu*, Anh Tran

CVPR, 2024

A high-quality dataset centered on extreme pose faces, supporting face synthesis, reenactment, recognition benchmarking, and more.

Project Page arXiv GitHub

2025. All rights Reserved. This website doesn't track you. Thanks to GIPHY for GIFs!