On December 26, 2024, our school invited two international scholars, Dr. Wang Yiming and Dr. Fabio Poiesi from Italy's FBK Institute, to deliver a successful academic seminar at the 1104 meeting room of the Feicui Science and Education Building (Building A). The seminar featured two presentations: "Addressing Vision Tasks with Large Foundation Models: How Far Can We Go Without Training" and "Understanding Point Clouds Through Geometric Descriptors and Natural Language." The seminar was hosted by Professor Zhong Zhun, with faculty and students from our school attending.

In her presentation, Dr. Wang Yiming discussed the latest advancements in Vision and Language Models (VLMs), emphasizing their generalization ability shown through large-scale network data pretraining. She also introduced how VLMs can be utilized for image classification and video understanding, providing new insights for resource-limited public research institutions.

Dr. Fabio Poiesi presented a new method for integrating geometric descriptors and natural language processing to understand point clouds, focusing on its applications in robot navigation and manipulation. He particularly discussed the method's performance in tasks such as point cloud registration, object pose estimation, functional segmentation, and dense descriptions.

Dr. Wang Yiming is a deep visual learning researcher at FBK, specializing in visual scene understanding, covering areas such as static modeling, semantic understanding, and video analysis. She obtained her Ph.D. from Queen Mary University of London in 2018 and previously worked at the Istituto Italiano di Tecnologia (IIT) on active 3D vision research. Dr. Wang serves as a reviewer for multiple conferences and is the associate editor for SoRo, managing low-carbon learning algorithm projects, and is a member of the ELLIS network.

Dr. Fabio Poiesi is the head of the Visual Technology Lab at FBK, where he joined in 2016 and became a tenured researcher in 2023. His research focuses on computer vision and deep learning, particularly in multi-object tracking, developing world-leading six-degree-of-freedom object pose estimation algorithms. He has won multiple BOP challenges and is also a member of the ELLIS network.

This seminar broadened the academic perspectives of our faculty and students, sparking their interest in deep visual learning and visual technology research. It also provided a platform for communication with international scholars. We look forward to more academic exchanges to further advance our research capabilities in this field.