Lei Zhang

Chair Professor of Computer Vision and Image Analysis

Fellow of IEEE
Department of Computing
The Hong Kong Polytechnic University
Hung Hom, Kowloon, Hong Kong

Office: PQ816
Email: cslzhang at comp.polyu dot edu.hk

I am also with OPPO Research Institute.

Education

3/1998~10/2001

PhD

Dept. of Automatic Control, Northwestern Polytechnical University, Xi'an, China.

9/1995~3/1998

M.Sc

Dept. of Automatic Control, Northwestern Polytechnical University, Xi'an, China.

9/1991~7/1995

B.Sc

Dept. of Aeronautical Engineering, Shenyang Inst. of Aeronautical Engineering, Shenyang, China.

Work Experience

7/2017~present

Chair Professor, Dept. of Computing, Hong Kong Polytechnic University, Hong Kong.

7/2015~6/2017

Professor, Dept. of Computing, Hong Kong Polytechnic University, Hong Kong.

9/2010~6/2015

Associate Professor, Dept. of Computing, Hong Kong Polytechnic University, Hong Kong.

1/2006~8/2010

Assistant Professor, Dept. of Computing, Hong Kong Polytechnic University, Hong Kong.

1/2003~1/2006

Postdoctoral Fellow, Dept. of Electrical and Computer Engineering, McMaster University, Canada.

1/2001~1/2003

Research Assistant/Associate, Dept. of Computing, Hong Kong Polytechnic University, Hong Kong.

Visual Computing Lab (our mission):

Y learning and beyond: for future visual enhancement and understanding.

 

My Google Scholar Citation Profile:

http://scholar.google.com/citations?user=tAK5l1IAAAAJ

http://t3.gstatic.com/images?q=tbn:ANd9GcSHajD6zIxvR7ORoWo3YUt1I4QtdrnCXbMSavwRvV19gHyDytAfYgMC900297235[1]

Papers&Codes

News

1.    Several PhD Student positions jointly trained with OPPO Research Institute are available. The research topics include Efficient LLM/VLM, Image/Video Restoration/Enhancement/Generation, Agent, etc., Please send me your CV if you have interest.

2.    Several Postdoctoral Fellow or Research Associate positions on Efficient LLM/VLM, Image/Video Restoration/Enhancement/Generation, Agent, etc., are available. Please send me your CV if you have interest.

3.    Research Interns on Efficient LLM/VLM, Agent, Image/Video Enhancement/Quality Assessment/Generation, etc., are available at OPPO Research Institute. Please send me your CV if you have interest.

Newly accepted

1.      N. Zhang, S. Wang, J. Zhang, C. Xiao, L. Zhang, "BitMTP: When Multi-Token Prediction Meets Low-Bit Large Language Models," in NeurIPS 2026. (paper) (code) (To make low-bit LLM faster: 2X speedup with MTP!)

2.      Z. Guo, R. Wu, Q. Yi, C. Xie, X. Wei, L. Zhang, " Two Halves are More than One: Phase-wise Velocity Distillation for Fast and High-Quality Image Generation," in NeurIPS 2026. (paper) (code) (Highly efficient and effective T2I with single-NFE equivalent inference cost!)

3.      Z. Zhang, L. Sun, R. Wu, Q. Yi, X. Kong, C. Xiao, L. Zhang, "LDM-is-AE: Latent Diffusion is an Auto-Encoder for End-to-End Image Generation," in NeurIPS 2026. (paper) (code) (LDM can do image generation by itself without encoder-decoder!)

4.      R. Cui, S. Liu, R. Wu, L. Zhang, "MESS: Multi-Exposure Sequence Synthesis for Generalizable Image Enhancement," in NeurIPS 2026. (paper) (code) (Synthesized data can train a generalized image enhancer!)

5.      P. Wang, S. Wang, L. Chen, Z. Ma, G. Zhang, L. Zhang, "DepthMaster: Unified Monocular Depth Estimation for Perspective and Panoramic Images," in NeurIPS 2026. (paper) (code) (New SOTA on perspective and panoramic images using a unified model!)

6.      C. Xie, Y. Wu, Q. Yi, L. Zhang, "Text-Vision Co-Instructed Image Editing," in NeurIPS 2026. (paper) (code) (Textual and visual prompts together make your editing more accurate!)

7.      G. Qin, J. Zhang, C. He, J. Liang, T. Wu, Y. Jin, L. Zhang, "Tool-IQA: Augmenting Image Quality Assessment with Simple Tools," in NeurIPS 2026. (paper) (code) (Magnifier and Gamma corrector are all tools you need to enhance your IQA model!)

8.      Y. Guo, Z. Zhang, P. Wang, X. Liang, Z. Ma, L. Zhang, "Memorize When Needed: Decoupled Memory Control for Spatially Consistent Long-Horizon Video Generation," in NeurIPS 2026. (paper) (code) (Efficient training for spatially consistent long-horizon video generation!)

9.      S. Liu, F. Han, J. Liao, Y. Sun, L. Sun, R. Li, X. Wei, J. Liang, H. Zeng, X. Zhang, L. Zhang, "AURA: An Autonomous Retouching Agent with Photographic Visual Thinking," in NeurIPS 2026. (paper) (code) (An agent to make your photo professional!)

10.  X. Kong, R. Wu, S. Liu, L. Sun, L. Zhang, "NSARM: Next-Scale Autoregressive Modeling for Robust Real-World Image Super-Resolution," in NeurIPS 2026. (paper) (code) (An efficient and robust AR model for real-world super-resolution!)

11.  Y. Zhong, H. Qin, X. Zhang, L. Zhang, G. Sun, "Breaking Modality Heterogeneity in Low-Bit Quantization for Large Vision-Language Models," in NeurIPS 2026. (paper) (code) (Joint quantization of vision and language modalities!)

12.  X. Kong, C. Dong and L. Zhang, "Towards Effective Multiple-in-One Image Restoration: A Sequential and Prompt Learning Strategy," in Machine Intelligence Research, 2026. (paper) (code) (A simple yet effective prompting strategy with a MiO-100 benchmark!)

Preprint

1.      X. Liu, S. Wang, X. Wei, M. Li, Z. Wang, Z. Yang, Z. Wen, L. Zhang, "OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding," preprint. (paper) (project page) (The first well-established benchmark for omni-modal video grounding and understanding!)

2.      X. Kong, S. Liu, R. Wu, L. Sun, Z. Zhang, J. Zhao, Y. Wu, L. Zhang, "HarnessIR: Harnessing Multimodal Foundation Models for Universal Real-World Image Restoration," preprint. (paper) (code) (MFMs are more powerful than you think for image restoration, depending on how you harness them!)

3.      Y. Yuan, W. Zhang, X. Li, S. Wang, K. Li, W. Li, J. Xiao, L. Zhang, B.C. Ooi, "PixelRefer: A Unified Framework for Spatio-Temporal Object Referring with Arbitrary Granularity," preprint. (paper) (code) (Spatio-temporal object referring at pixel level!)

4.      Z. Zhao, Q. Xie, K. Huang, L. Zhang, D. Meng, and Z. Xu, "Rotation Equivariant Mamba for Vision Tasks," preprint. (paper) (code) (A mamba network with rotation invariant property!)

5.      Z. Zheng, C. He, J. Zhang, M-M. Cheng, L. Zhang, "HaloBlock: Faster and Better Decoding for Diffusion Large Language Models," preprint. (paper) (code) (Training-free method to accelerate dLLM decoding!)

6.      L. Sun, R. Wu, X. Kong, J. Zhao, Q. Yi, Y. Sun, S. Liu, Z. Zhang, L. Zhang, "PixRestore: Unified Image Restoration via Pixel Diffusion Transformer," preprint. (paper) (code) (Pixel diffusion provides an effective and efficient solution for unified image restoration!)

7.      R. Li, T. Yang, Z. Ma, F. Ai, S. Wen, L. Zhang, "Avatar-Forever: Decoupled Parallel Training for High-Quality Real-Time Infinite Avatars," preprint. (paper) (code) (Endless high-quality avatar video generation in real time!)

8.      W. Zhu, Y. Zhang, W. Zeng, L. Zhang, "MMOOC: A Comprehensive Benchmark for Out-of-Context Evaluation in Multimodal Large Language Models," preprint. (paper) (code&data)

9.      X. Kong, J. Zhao, L. Sun, R. Wu, L. Zhang, "GGT-100K: Generative Ground Truth for Generalizable Real-World Image Restoration," preprint. (paper) (code) (Can multimodal foundation models be the solution for generalizable real-world image restoration?)

10.   R. Li, T. Yang, F. Ai, T. Wu, S. Wen, B. Peng, Lei Zhang, "Long-Horizon Streaming Video Generation via Hybrid Attention with Decoupled Distillation," preprint. (paper) (code) (Video generation at 29.5 FPS (832X480) on a single H100 GPU without quantization or model compression!)

11.   W. Li, Z. Qi, Z. Zhao, K. Zhang, L. Zhang, "Weighted Reverse Convolution for Feature Upsampling," preprint. (paper) (code) (Making the features of vision foundation models stronger!)

12.   Z. Zheng, C. He, S. Wang, Y. Li, M. Cheng, L. Zhang, "DEL: Digit Entropy Loss for Numerical Learning of Large Language Models," preprint. (paper) (code) (A simple yet effective loss to improve the numerical learning capability of LLMs!)

13.   H. Wang, C. Shen, L. Zhang, Z. Cheng, "ATSS: Detecting AI-Generated Videos via Anomalous Temporal Self-Similarity," preprint. (paper) (code) (A highly effective algorithm to detect AI-generated videos!)

14.   J. Zhang, C. Xiao, A. Wu, X. Zhang, L. Zhang, "Pretraining A Large Language Model using Distributed GPUs: A Memory-Efficient Decentralized Paradigm," preprint. (paper) (code) (Can we train large-scale LLMs using GPUs with low memory? )

15.   Z. Wang, X. Wei, B. Li, Z. Guo, J. Zhang, H. Wei, K. Wang, L. Zhang, "VideoVerse: Does Your T2V Generator Have World Model Capability to Synthesize Videos?" preprint. (paper) (code) (To evaluate how strong your T2V model is!)