[Intern position]: I am hiring self-motivated graduate/undergraduate students. Intern positions are always available. Please email me with your CV if you are interested.
NEW (08/2024): I will be serving as the Associate Editor of IEEE Transactions on Image Processing.
NEW (08/2024): I was elected as a senior member of IEEE.
NEW (06/2024): Our research on generative foundational models for remote sensing was reported by the National Natural Science Foundation of China (NSFC).
NEW (12/2023): I am organizing a special issue for the new Nature journal, Communications Engineering, which is the first Nature journal dedicated to the engineering field. The theme is “Technologies for Augmented and Virtual Reality.” Submissions are welcome.
NEW (04/2023): Our paper on driving environment simulation got accepted to Nature Communications.
NEW (03/2023): Our paper on autonomous vehicle testing got accepted to Nature (cover). Congratulations!
NEW (02/2023): One paper accepted to CVPR 2023 (featured in New Scientist (新科学人)).
NEW (01/2023): Our survey paper on object detection got accepted by Proceedings of the IEEE.
Dr. Zhengxia Zou's primary research interests include computer vision and deep learning (e.g., visual perception, generation, and predictive modeling), with applications in remote sensing and autonomous driving. He has published papers as the first/corresponding author in leading journals such as Proceedings of the IEEE, Nature Communications, IEEE Transactions on Pattern Analysis and Machine Intelligence, IEEE Transactions on Image Processing, and IEEE Transactions on Geoscience and Remote Sensing, as well as in major conferences like IEEE/CVF Computer Vision and Pattern Recognition Conference and IEEE/CVF International Conference on Computer Vision. He has co-authored over 80 academic papers, including Nature (cover), with five highly cited and three hot papers in ESI. His work has been cited over 7000 times on Google Scholar, with one paper receiving over 3000 citations. His open-source projects on GitHub have over 4000 subscriptions, and two of his papers were selected as “Top-10 Trending Research” on PapersWithCode. He was listed among the top 2% of scientists worldwide in 2022 and 2023 (Stanford/Elsevier’s Top 2% Scientists). He is an Associate Editor of IEEE Transactions on Image Processing. His research has been reported by media outlets such as Xinhua News Agency, NSFC, New Scientists, AAAS, MIT Technology Review, and etc.
Zhengxia Zou* (*Corresponding author), Zhenwei Shi, Yuhong Guo, and Jieping Ye*. Object Detection in 20 Years: A Survey. Proceedings of the IEEE, Volumn 111, Issue 13, 2023. [2600+ citations], Ranks #1 in Most Popular Articles, Ranks #1 in Most Downloaded Articles. [PDF]
Extensively reviews the fast-moving research field of object detection in the light of technical evolution, spanning over a quarter-century's time (from the 1990s to 2022). A number of topics are covered in this paper, including milestone detectors, detection datasets, metrics, fundamental building blocks of detection systems, speed up techniques, and the recent state of the art detection methods.
- High-impact citation: Stanford CS231n: Convolutional Neural Networks for Visual Recognition (2019-2023).
- Media coverage: [PIEEE] Proceedings of the IEEE Paper Explores Object Detection's Evolution and Future Directions | [专知] 密歇根大学40页最新论文带你全面了解目标检测
- In other languages:  English-to-Chinese (1) | English-to-Chinese (2)
Zhengxia Zou, Rusheng Zhang, Shengyin Shen, Gaurav Pandey, Punarjay Chakravarty, Armin Parchami, and Henry X. Liu (*Corresponding author). Real-time Full-stack Traffic Scene Perception for Autonomous Driving with Roadside Cameras. The International Conference on Robotics and Automation (ICRA), 2022. [PDF] [1min-DemoVideo]
Deployed at a two-lane roundabout located at Ellsworth Rd. and State St., Ann Arbor, MI, USA, providing 7x24 real-time traffic flow monitoring for hazardous driving scenarios identification.
- Media coverage: [Mcity] Mcity 2.0 Mobility Data Center.
Zhiping Yu, Chenyang Liu, Liqin Liu, Zhenwei Shi, Zhengxia Zou* (*Corresponding author). MetaEarth: A Generative Foundation Model for Global-Scale Remote Sensing Image Generation. ArXiv, 2024. [PDF]
- Media coverage: [国家自然科学基金委] 我国学者在遥感生成式模型方向取得进展 | [北京航空航天大学] 用AI俯瞰全球,北航最新研究成果 | [中国图象图形学学会] [量子位] 把整个地球装进神经网络,北航团队推出全球遥感图像生成模型
Rui Zhao, Wei Li, Zhipeng Hu, Lincheng Li*, Zhengxia Zou* (*Corresponding author), Zhenwei Shi, Changjie Fan. Zero-Shot Text-to-Parameter Translation for Game Character Auto-Creation. Accepted to IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2023. [PDF]
- Featured apps: Justice Mobile (The first plausible solution for text-driven game characters auto-creation).
- Media coverage: [New Scientist (新科学人)] Character creator AI puts Barack Obama – or anyone – in a video game | [AUTOMATON] AI利用で「言葉で頼むだけでキャラクリしてくれる」技術成果を研究者らが報告。動物から著名人まで幅広くテキストで指定可
Zhengxia Zou (*Corresponding author), Rui Zhao, Tianyang Shi, Shuang Qiu, and Zhenwei Shi. Castle in the Sky: Dynamic Sky Replacement and Harmonization in Videos. IEEE Transactions on Image Processing, 2022. In press. [PDF] [Project] [GitHub (2.0k☆)]
- Media coverage: [机器之心] 建造自己的“天空之城”,密歇根大学博士后的这项研究可以虚空造物、偷天换日 | [TNW] This open-source AI tool can make your video spectacular with sky replacement effects | [Weights and Biases] The Sky Is In Our Grasp
Zhengxia Zou (*Corresponding author), Tianyang Shi, Shuang Qiu, Yi Yuan, and Zhenwei Shi. Stylized Neural Painting. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2021, Oral Presentation.[PDF][Project] [GitHub (1.5k☆)]
- Featured apps: REMINI and 你我当年, two photo editors with 10M+ users worldwide | RunwayML, a web-based video editor
- Media coverage: [新华社] 动动手,一起为春天中国“添彩” - 送您一支AI画笔, 为祖国春天涂抹万千风情 | [中国记者(新华社期刊)] 风格化神经绘画技术在主题宣传中的应用 | [机器之心] 有了这支矢量神经风格画笔,无需GAN也可生成精美绘画 | [MarkTechPost] An Image-To-Painting Translation Method That Generates Painting Artworks With Controllable Styles
Xintao Yan+, Zhengxia Zou+ (+equal contribution), Shuo Feng, Haojie Zhu, Haowei Sun, Henry X. Liu* (*Corresponding author). Learning Naturalistic Driving Environment with Statistical Realism. Nature Communications (featured image), 2023. Featured in Editors' Highlights.[PDF][Github]
We develop NeuralNDE, a Transformer-based framework to learn multi-agent interaction behavior from real-world vehicle trajectory data. NeuralNDE can achieve both accurate safety-critical and normal driving statistics.
- Media coverage: [EurekAlert] [TechXplore] [Michigan News] World’s first realistic simulated driving environment based on ‘crash-prone’ Michigan intersection. | [量子位] 自动驾驶仿真系统登Nature子刊,准确建模事故率事故类型,全华人团队打造 | [MIT科技评论] 密西根大学团队开发高精度自然驾驶仿真算法,为自动驾驶开发提供仿真基础
Shuo Feng, Haowei Sun, Xintao Yan, Haojie Zhu, Zhengxia Zou, Shengyin Shen, Henry X. Liu* (*Corresponding author). Dense Reinforcement Learning for Safety Validation of Autonomous Vehicles. Nature (cover), 2023. [PDF][Github]
One critical bottleneck that impedes the development and deployment of autonomous vehicles is the prohibitively high economic and time costs required to validate their safety in a naturalistic driving environment, owing to the rarity of safety-critical events. Here we report the development of an intelligent testing environment, where artificial-intelligence-based background agents are trained to validate the safety performances of autonomous vehicles in an accelerated mode, without loss of unbiasedness.
- Media coverage: [Nature News] Hazards help autonomous cars to drive safely. | [Nature Podcast] How to make driverless cars safer — expose them to lots of dangerous drivers | [Michigan Engineering] Simulated terrible drivers cut the time and cost of AV testing by a factor of one thousand