Multi-view X-ray Image Synthesis with Multiple Domain Disentanglement from CT Scans
ACM MM 2024 · Accepted Jul 2024 CCF A
Disentangles anatomy and image style to synthesize multi-view X-rays from CT scans without paired real X-ray training data.
I am a Ph.D. student in Computer Science at Beihang University’s State Key Laboratory of Virtual Reality Technology and Systems, jointly advised by Associate Researcher Qing Xia and Prof. Aimin Hao. I hold a master’s degree from the University of Science and Technology Beijing and a bachelor’s degree from Guizhou University.
I am currently a Research Intern with Canva’s AI Efficiency team, where I contribute to Canva Craft, an image generation model that debuted at No. 4 on DesignArena’s image leaderboard (the top three were all GPT-Image series models). Previously, I interned with the human body team at Meitu’s MT Lab, working on diffusion models and personalized image generation.
My research interests include cross-modal representation alignment between motion and language, generative artificial intelligence, and medical image processing. Recently, I have been collaborating closely with Dr. Yang-Hao Zhou on video generation and understanding.
I have served as a reviewer for Future Generation Computer Systems (FGCS), as recorded on my ORCID profile.
I welcome research collaborations—feel free to get in touch!
* Corresponding author · † Equal contribution (co-first authors) · ‡ Project Leader
ACM MM 2024 · Accepted Jul 2024 CCF A
Disentangles anatomy and image style to synthesize multi-view X-rays from CT scans without paired real X-ray training data.
BIBM 2026 · Accepted Sep 2026 CCF B
* Corresponding author · † Equal contribution (co-first authors) · ‡ Project Leader
Preprint · First posted Oct 2026
Plans causal physical state transitions with multiple agents to improve video consistency through prompt enhancement, without retraining the generator.
Preprint · First posted Oct 2026
Prunes and merges redundant video patches before vision encoding to reduce computation without additional training.
Preprint · First posted Sep 2026
Tests whether motion-language evaluators ground temporal order, mirror reflection, and action identity beyond text-only shortcuts.
Preprint · First posted Sep 2026
Evaluates generated medical teaching videos against clinical action checklists, separating visual realism from procedural correctness.
Preprint · First posted Jul 2024
Combines edge attention and frequency enhancement to reconstruct vertebral geometry from two X-ray views.
Ph.D. in Computer Science · In progress
Sep 2025 – Present
M.Eng. in Electronic Information
MedAI Lab · Advised by Dr. Shuang Song and Prof. Ruoxiu Xiao
Sep 2022 – Jun 2025
B.Eng. in Communication Engineering
Research mentors: Prof. Hanyu Lu and Prof. Rongfen Zhang
Sep 2017 – Jun 2021
Research Intern
Distillation · Reinforcement Learning · Agents
Apr 2026 – Present
Canva Craft · Contributor
I contributed to Canva Craft during my research internship at Canva.
Official Canva announcement · Oct 22, 2026
View original post on X
Computer Vision Engineer (Intern)
Diffusion Models · FLUX · Personalized Generation
Jun 2025 – Sep 2025
Embedded Software Engineer
Embedded C Programming · C# Software Development
Jul 2021 – Sep 2021
Outside research, I enjoy badminton, jump rope, and reading.
I enjoy films and especially recommend The Secret Life of Walter Mitty.
I also enjoy listening to vaporwave.