4月29日上午11:00,我院邰颖老师邀请密歇根州立大学xiaoming Liu教授作题为“Visual Recognition in the Era of Foundation Models”的学术报告。
报告摘要:
Visual recognition aims to recognize objects, or their instances given an imagery. This is one of the most fundamental tasks that computer vision researchers are striving to solve in the past decades. With the recent emerging of large foundation models, being LLM or VLM, researchers are actively studying how to advance visual recognition in the era of foundation models. Some common questions include 1) how to leverage pre-trained foundation models for visual recognition? 2) how to continue the innovation of vision transformer in light of downstream tasks? 3) how to design and learn our own foundation model for a specific task? This talk will shed some lights on how we have been answering these questions in recent years. I will also share some of our recent works on 3D vision and anti-deepfakes.
报告人介绍:
Dr. Xiaoming Liu现任密歇根州立大学(MSU)计算机科学与工程系的MSU Foundation Professor,以及Anil and Nandita Jain冠名教授。他于2004年获得卡内基梅隆大学博士学位,在2012年加入密歇根州立大学之前,曾担任通用电气(GE)全球研究院的研究科学家。他的研究方向涵盖计算机视觉、机器学习和生物识别技术,尤其专注于人脸相关分析与三维视觉领域。自2012年起,他助力密歇根州立大学构建了实力强劲的计算机视觉学科,根据csrankings.org的五年统计数据显示,该校该学科已跻身全美前15名。他将担任2028年国际计算机视觉与模式识别会议(CVPR)的联合程序主席,并担任IEEE Transactions on Pattern Analysis and Machine Intelligence的副主编。他已发表200余篇学术论文,提交35项美国专利申请,其研究成果被谷歌学术引用超过3万次,H指数达81。刘晓明博士是电气与电子工程师学会(IEEE)和国际模式识别协会(IAPR)的会士。
文字 黄旭鸿
图片 汤靖玲