At the recently held 2023 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Prof. Yu Hongshan and his research team won the first prize for their excellent performance in the AQTC (Affordance-Centric Question-driven Task Completion Challenge) in the LOVEU (Long-form VidEo Understanding and Generation) competition, and were invited to give an oral presentation at the workshop.

Prof. Yu and his research team are from HNU College of Electrical and Information Engineering.

HNU and the University of Central Florida collaborated for the competition. Four HNU students participated in the competition: Chen Tongjia and Miao Wei, postgraduate students of grade 2021; Li Zechuan, a doctoral candidate of grade 2021; and, Wang Jingwen, a doctoral candidate of grade 2020, and mentored by Prof. Yu Hongshan, Prof. Sun Wei, and Dr. Yang Zhengeng.

In the competition, Prof. Yu's team proposed a function-interaction centric approach with spatiotemporal visual-language alignment. This method successfully addressed two challenges: the alignment of multimodal data collaborative analysis and the previous methods' weak perception of interactive information. This project is expected to be applied to such practical scenarios as first-person AR assistant and VR glasses in the future.

Organized by IEEE, CVPR is one of the top-tier international conferences of the highest influence and significance in the computer vision and pattern recognition field, and is also a Class-A international conference recommended by the China Computer Federation (CCF).

The CVPR 2023 LOVEU competition was co-organized by the National University of Singapore, MetaAI, University of Technology Sydney, University of California, Berkeley, and Texas A&M University. A number of famous universities and research institutions from all over the world participated.

A group photo of Prof. Yu and his research team.