Chai Wenhao

23’ Civil Engineering undergraduate of ZJUI

Aiming and Positioning, Riding the Wind and Waves

Challenge without Limits

Chai Wenhao has already established a clear plan for his future development and devotes himself to the field he loves with practical actions. "I have tried many different research directions to find the field I am interested in. Through the research experience accumulated in SRTP and communication with professors in ZJUI and the research team, I have gradually clarified my research interests and decided to focus on the field where I am interested." He has firmly established his path of research and exploration and has invested a lot of time and energy in research, maintaining a long-term research work schedule of ten hours per day, never slacking off. Moreover, he actively participated in lab projects and delved into experimental techniques and data analysis methods. “Research is not just theory, but practice and experimentation are also indispensable parts of research. Through practical operations, I am able to apply theoretical knowledge to practical problems and verify my research hypotheses.”

In addition to research training, he also actively participated in academic conferences and exchange activities and offered to exchange and share research results with other researchers in the field. In June 2022, Chai Wenhao attended the CVPR2022 Academic Conference in New Orleans, the United States, which is the world's top academic conference in the field of Artificial Intelligence and Machine Learning. The impact factor is second only to Science and Nature. “Participating in academic conferences has broadened my horizons and allowed me to have the opportunity to communicate with scholars all around the world and establish many connections with them, " he shared. During his undergraduate studies, Chai Wenhao officially published 9 academic papers, and currently 10 papers are under peer review. Among the published and reviewed papers, 10 were those of which he was the first author or co-first author.

Regarding life, Chai Wenhao is enthusiastic and resilient, unrestrained and calm. In his spare time, he chose to briefly escape from his studies and work, travel through the world, and return to his purest self in love. During the six months of UIUC exchange, he traveled through Chicago, Boston, New York, Orlando, Miami, New Orleans, Seattle, Los Angeles, and other cities to experience different cultural atmospheres and environments. “I believe that a person's thoughts can be sublimated through cultural communication and collision, which is also a unique advantage given by ZJUI to every student." He, who loves photography, is accompanied by the light and shadow all the way, capturing the liveliness and beauty along the way.

▲ Chai Wenhao traveled through U.S. during his UIUC exchange

Yearning for the Love in Heart, Breaking Through Boundaries

No Limits in Life

"I have always had a strong interest in the field of Computer Vision, and I believe it will play an important role in the future. I hope to combine the fields of Civil Engineering and Computer Science." Chai Wenhao believes that ZJUI's interdisciplinary education and research platform breaks through the limitations of traditional disciplines, broadens his horizons, expands his skills, strengthens his belief in trying to cross professional boundaries, and helps him bravely pursue his interests and applies them to practical research.

At the beginning of his attempt to conduct interdisciplinary research on Computer Vision, he was worried that he would not be able to find a suitable research group due to the lack of relevant basic knowledge. Fortunately, he actively utilized the cross-disciplinary platforms provided by ZJUI and the Student Research and Training Program (SRTP), and successfully joined computer related research projects and participated in training and seminars related to Computer Vision to expand his knowledge in width and depth. In addition, he also made full use of his spare time to carry out internship practice. He once interned in the National Center for Supercomputer Applications of the United States and the Microsoft Research Asia. Through continuous self-learning and practice, he gradually overcame the difficulties brought by the lack of basic knowledge and made progress in the field of Computer Vision. After successfully publishing his first paper in his junior year, he had more confidence and motivation to conduct in-depth research and bravely climb to new heights. In the next two years, Chai Wenhao achieved numerous achievements in research and won multiple awards and honors.

Recently, as an independent first author, he has published two high-level papers at the top conference ICCV 2023 on Computer Vision and Artificial Intelligence. In the article 'Global Adaptation meetings Local Generalization: Unsupervised Domain Adaptation for 3D Human Pose Estimation', he achieved significant performance improvement by combining a simple and effective unsupervised domain adaptation framework of global adaptation and local generalization in PoseDA for 3D human pose estimation, without introducing additional learnable parameters. Meanwhile, this paper proposes the Local Pose Amplification (LPA) module to enhance the diversity of 3D poses through the confrontation training scheme. This method can be applied to almost all 2D-3D Lifting models. Under the cross-dataset evaluation setting of MPI-INF-3DHP, PoseDA achieved a 61.3mm MPJPE, which is 10.2% higher than the previous optimal method. In the article “StableVideo: Text driven Consistency-aware Diffusion Video Editing”, he introduces temporal dependency to existing text-driven diffusion models, which allows them to generate consistent appearance for the new objects. Specifically, he develops a novel inter-frame propagation mechanism for diffusion video editing, which leverages the concept of layered representations to propagate the geometry and appearance information from one frame to the next.

▲Global Adaptation meets Local Generalization: Unsupervised Domain Adaptation for 3D Human Pose Estimation论文图表

"ZJUI's superior cross-disciplinary platforms, excellent active learning ability, and active cooperation with instructors and peers are the key for me to simultaneously complete Civil Engineering studies and participate in research in the field of Computer Science during my undergraduate studies. I hope to integrate knowledge from both Civil Engineering and Computer Vision to contribute to solving practical problems," said Chai Wenhao. Currently, he is about to pursue a master's degree in EE (Electrical Engineering) at the University of Washington, focusing on Computer Vision and General Intelligence.

On the Way of Crossing Boundaries, Guided by Bright Lights

Platform with Infinite Width

"ZJUI's international cooperative education cultivation model and environment have provided us with rich opportunities for overseas exchanges, and the English-speaking environment has also paved the way for my research and exchange. More importantly, ZJUI's interdisciplinary convergence and integration provides a variety of choices for undergraduate research.” Although Chai Wenhao is an undergraduate student of Civil Engineering, he joined ZJUI Assistant Prof. Wang Gao’ang's CVNext Lab due to his strong interest in Computer Vision.

If Chai Wenhao's outstanding achievements in research are due to ZJUI's cross- disciplinary cultivation characteristics and the "One Thousand and One Nights" efforts in the laboratory, then this "One Thousand and One Nights" must be inseparable from his instructor, Assistant Prof. Wang Gao’ang of ZJUI. In the words of Chai Wenhao, Assistant Prof. Wang Gao’ang is an experienced and passionate instructor who led him to move forward without hesitation on the road of research.

"Prof. Wang Gao’ang is dutiful since he will arrange one-to-one exchange meetings for each student. From the initial idea discussion to the final paper writing, Prof. Wang has given us very detailed guidance. In addition, our group will organize academic seminars every week to share research progress and experience, so that we can broaden our research vision and experience research taste. I feel lucky to be guided by Prof. Wang Gao’ang, who not only supports me in academic, but also has taught me how to train scientific thinking and independent research abilities.” On the path of research, Assistant Prof. Wang Gao’ang is his instructor and his helpful friend.

"Wen Hao has a great passion for research. He invests a lot of time and energy in literature reading and code implementation every day. He always pays attention to the forefront of technology and has a profound understanding of the challenges in the field of Artificial Intelligence. I think this is very valuable, "commented by Assistant Prof. Wang Gao’ang.

The path of research is long, with guidance from good instructors and peers, mutual encouragement and stimulation, he can go further and be more confident.

Suggestions for ZJUIers

1. Motivation is all you need. A strong motivation is the driving force behind research and learning. This motivation can be for academic application, subsequent employment, or interest exploration. When we find the motivation for research, we can truly spend time and produce results. Although the path of research may not be suitable for everyone, daring to try will definitely yield your own rewards.

2. Maintain a balance between research, academic performance, and daily life. Investing time and effort in research while ensuring GPA is a relatively safe choice; At the same time, appropriate relaxation and entertainment are beneficial for maintaining physical and mental health. In addition, it is necessary to balance research experiments with reading literature, actively verify theories in experiments, and read extensively and deeply. However, learning without thinking is useless, and thinking without learning is dangerous. A reasonable allocation should be made between the two.

3.Publishing the first paper is the key to building research confidence. In the fields of Computer Vision and Deep Learning, it is not uncommon for undergraduate students to be able to publish papers at top conferences. Actively participating in projects and publishing one's first paper is crucial for building academic confidence in research. Only with confidence can we reap more results.