Dawei Guan

I'm a master's student at the School of Artificial Intelligence and Data Science, University of Science and Technology of China (USTC), supervised by Prof. Jiangtao Wang and co-supervised by Assoc. Prof. Di Yang.

My research interests include motion generation, multimodal generative models, and 3D scene reconstruction and editing.

I received my bachelor's degree from the School of Electronic and Information Engineering, Beijing Jiaotong University (BJTU).

Research

Harnessing Coupled Stream Completion for Human-Object Interaction Modeling

Dawei Guan, Di Yang, and Jiangtao Wang

arXiv, 2026

arXiv project page

TRACE keeps body, object, and hand motion in separate latent streams while coupling their updates to maintain coherent interactions. The same model supports completion of any missing stream and text-based interaction understanding.

Overview of the FlowCoMotion text-to-motion generation framework

FlowCoMotion: Text-to-Motion Generation via Token-Latent Flow Modeling

Dawei Guan, Di Yang, Chengjie Jin, and Jiangtao Wang

arXiv, 2026

arXiv project pageCode coming soon

We couple discrete motion tokens with continuous latents to capture both high-level semantics and fine motion details. A text-conditioned flow model then generates human motion from this unified representation.

Overview of the DaptDiffusion interactive image editing framework

DaptDiffusion: Enhancing Pixel-level Interactive Editing with Dense-UNet and Adam Point Update in Diffusion Models

Dawei Guan and Wei Li

Applied Intelligence, 2025

paper code

DaptDiffusion makes drag-based image editing more efficient and robust by combining a Dense-UNet feature network with an Adam-inspired point update rule in the latent space of diffusion models.

Overview of the VariGAN image style transfer framework

VariGAN: Enhancing Image Style Transfer via UNet Generator, Depthwise Discriminator, and LPIPS Loss in Adversarial Learning Framework

Dawei Guan, Xinping Lin, Haoyi Zhang, and Hang Zhou

Sensors, 2025

paper code

VariGAN improves unpaired image style transfer with an attention-enhanced UNet generator, a depthwise discriminator, and LPIPS perceptual loss, refining both the training process and the visual quality of transferred images.

News

2026.04We released FlowCoMotion on arXiv. Visit the project page for details.
2026.01I launched my academic homepage.

Education

2026.09 – present
University of Science and Technology of China
Master's student · Hefei, China
2022.09 – 2026.06
Beijing Jiaotong University
Bachelor's degree · Beijing, China

Acknowledgement

I'm especially grateful to my girlfriend, Yitong Liu, who provided the profile picture for this page and is a constant source of motivation on my journey.