The 42nd Computer Graphics International (CGI 2025), hosted by The Hong Kong Polytechnic University from July 14 to 18, is a Category C international academic conference recommended by the China Computer Federation (CCF). Zhou Guiren and Ma Xin, undergraduate students from SCMU's School of Computer Science (School of Artificial Intelligence), made oral presentations as first authors on their research about intelligent recognition of Nvshu script and digital preservation of traditional ethnic sports respectively. This achievement marks a breakthrough for SCMU undergraduates presenting research at high-level international academic conferences.

Zhou Guiren and Ma Xin at CGI 2025. Photo by Zhu Jianlin

Zhou Guiren Delivering Oral Presentation. Photo by Zhu Jianlin

Ma Xin Delivering Oral Presentation. Photo by Zhu Jianlin

Zhou Guiren's paper titled "Enhancing Nvshu Recognition Based on Polarity-aware Linear Attention and Learnable Local Salient Kernel" addresses the digital preservation challenges of Nvshu, an endangered writing system. The proposed PALLS model resolves issues of stroke adhesion and sample scarcity through polarity-aware attention mechanisms and learnable local saliency kernel modules.

Ma Xin and her team's paper "An Improved Small Object Detection Method for Shuttlecock Activity Analysis" focuses on the traditional ethnic sport of shuttlecock. By improving the YOLO framework, they solved high-speed tiny object detection challenges and developed a web-based analytical system for real-time motion visualization, overcoming limitations of manual recording and advancing scientific training methods for ethnic sports.