Zhenhua Yang
My name is Zhenhua Yang (杨振华, Yeung Chenwa).
I am a Full-time Algorithm Engineer in Taobao & Tmall Group of Alibaba (2025). I received my Master's and Bachelor's degree from SCUT-DLVCLab [supervised by Prof. Lianwen Jin] in South China University of Technology.
Fortunately, I have interned at the Kling Team (advised by Xin Tao) and IDEA, Shenzhen (advised by Lei Zhang and Hao Zhang).
My research interests are focused on:
- Post-Training for Multi-modal Generation and Editing
- Unifying Understanding and Generation for Next Generation Model
- Visual Text Generation and Editting (Text Rendering)
News
- 05/2025 🎉🎉🎉 Our work AutoHDR is accepted by ACL 2025 Main. I am the project leader. Code released.
- 02/2025 🎉🎉🎉 Our paper HDR is selected as the Oral presentation.
- 12/2024 The inference code of our paper HDR is released at GitHub.
- 12/2024 Our paper HDR is accepted by AAAI 2025 🎉🎉🎉.
- 07/2024 I will attend the ICML 2024 conference in person in Vienna, Austria.
- 06/2024 Interning at IDEA, supervised by Prof. Lei Zhang, working on vision-language large models for video understanding.
- 05/2024 Our paper UPOCR is accepted by ICML 2024 🎉🎉🎉.
- 12/2023 🔥🔥🔥 The Hugging Face Demo and GitHub of FontDiffuser are released!
- 12/2023 🎉 Paper FontDiffuser is accepted by AAAI 2024.
- 12/2023 Our paper UPOCR is released to arXiv.
Education
South China University of Technology
Sep. 2022 - 2025
M.S. Student at SCUT-DLVCLab, School of Electronic and Information Engineering
South China University of Technology
Sep. 2018 - Jun. 2022
B.E. Student in School of Automation Science and Engineering
Experience
Alibaba - Taobao & Tmall Group
July 2025 - Present
Full-time AIGC Algorithm Engineer
Post-Training for Image Generation and Editing.
KuaiShou - Kling Team
Jan. 2025 - Apr. 2025
Research Intern, supervised by Xin Tao.
Topic: Unifying Understanding and Generation for Next Generation Model.
International Digital Economy Academy (IDEA) - CVR
Jun. 2024 - Sep. 2024
Research Intern, supervised by Prof. Lei Zhang.
Topic: MLLM for Streaming Video Narration and QA.
Selected Publications [Full List]
Predicting the Original Appearance of Damaged Historical Documents
Zhenhua Yang, Dezhi Peng, Yongxin Shi, Yuyi Zhang, Chongyu Liu, Lianwen Jin†
Proc. of the AAAI Conference on Artificial Intelligence (AAAI Oral), 2025
FontDiffuser: One-Shot Font Generation via Denoising Diffusion with Multi-Scale Content Aggregation and Style Contrastive Learning
Zhenhua Yang, Dezhi Peng, Yuxin Kong, Yuyi Zhang, Cong Yao, Lianwen Jin†
Proc. of the AAAI Conference on Artificial Intelligence (AAAI), 2024
UPOCR: Towards Unified Pixel-Level OCR Interface
Dezhi Peng, Zhenhua Yang, Jiaxin Zhang, Chongyu Liu, Yongxin Shi, Lianwen Jin†
International Conference on Machine Learning (ICML), 2024
Reviving Cultural Heritage: A Novel Approach for Comprehensive Historical Document Restoration
Yuyi Zhang, Peirong Zhang, Zhenhua Yang* (Project Lead), et al., Lianwen Jin†
Meeting of the Association for Computational Linguistics (ACL Main), 2025
Draft, Verify, Restore: Self-Refining Historical Inscription Restoration with a Unified MLLM
Topic: Unify Content Reasoning (Understanding) and Generation
…, Zhenhua Yang* (Project Lead), et al., Lianwen Jin†
Submitted to Meeting of the Association for Computational Linguistics (ACL), 2026
Open-Source Projects
Optical Character Recognition with Segment Anything (OCR-SAM)
Focus: Exploring SAM's zero-shot generalization in OCR tasks. Integrated SAM with MMOCR to develop specialized application demos, including precise text removal and high-fidelity text inpainting with a Gradio-based WebUI.
Flexible Diffusion-based Font Generation Framework
Focus: A robust generative framework for few-shot font stylization. Capable of synthesizing unseen characters and complex styles, supporting cross-lingual generation (e.g., Chinese to Korean) via advanced diffusion denoising.
Recommendations of Diffusion for Text-Image
Focus: A curated academic collection of state-of-the-art diffusion models for text-centric visual tasks, including visual text generation, font synthesis, and scene text recognition.
Recommendations of Document Image Processing
Focus: A comprehensive survey of document image restoration techniques, covering appearance enhancement, deshadowing, dewarping, deblurring, and binarization.
Awards
- Shenzhen HighPower Technology Scholarship, 2022 (Top 2%)
- First-Class Campus Scholarship, 2021 (Top 5%)
- American Mathematical Contest in Modeling, Meritorious Prize, 2020
- Alibaba Tianchi Competition, Top 1.2%, 2021
Mics
Hobby: 🎣 Fishing, 🏊♂️ Swimming, 🚲 Riding, 🏓 Table tennis, 🏀 Basketball, 🎤 Singing. Learning 🎹 Piano.
Languages: Chinese (Mandarin, Cantonese, Hakka), English.
Habit: A heavy coffee drinker ☕.