報告題目：Machine Learning Solutions to Visual Recognition \r
Problems
主講嘉賓：華揚助理教授 英國貝爾法斯特女王大學 ( Queen’s University, Belfast, \r
UK）
邀請人：鄒文斌老師
時間：2017年6月29日（周四）上午10:10 -- \r
11:00
地點：深圳大學南區(qū)基礎實驗樓北座信息工程學院N710會議室
報告摘要:
One of the ultimate \r
goals of computer vision is to enable computers to duplicate the main functions \r
of human vision or even work better under certain conditions. However, human \r
vision is far more complex than it looks like. In the last decade, computer \r
vision, especially in the field of visual recognition, has been developed \r
prosperously, mostly driven by machine learning technology. On one side, machine \r
learning formulates diverse computer vision problems as proper mathematical \r
models and solves them efficiently. On the other side, computer vision rooting \r
in real-world scenarios serves as a great testbed to evaluate the validity and \r
the efficiency of machine learning methods.
In this talk, I will briefly \r
review my research activities on different novel machine learning solutions for \r
real-world visual recognition problems, specifically, kernel method for general \r
object classification and detection, energy minimization for long-term object \r
tracking and deep learning for face recognition.
嘉賓簡介：
Yang Hua will \r
start as a Lecturer (equivalent to Assistant Professor) at Queen’s University \r
Belfast, UK in July 2017. He obtained his Ph.D. under the supervision of Dr. \r
Cordelia Schmid and Dr. Karteek Alahari in Inria LEAR/THOTH team, France. His \r
Ph.D. research was funded by the scholarship of the Microsoft Research - Inria \r
joint center.
Yang Hua is now a research scientist at AnyVision (UK \r
office). Before his Ph.D. study, he worked as a senior R&D engineer at \r
Panasonic R&D Center Singapore from 2008 to 2012, where he was engaged in a \r
long-term research project and the related industrial product development. \r
With more than nine-year combined experience of research and engineering in \r
computer vision and machine learning, he published several first-tier \r
international journal articles and conference papers with more than 460 \r
citations, he holds/co-holds three US patents and one China patent, he designed \r
and developed fully optimized image classification module for Panasonic’s new \r
product line launched in Japan, and he won four titles of prestigious \r
international competitions in computer vision spanning image and video domains, \r
including the PASCAL Visual Object Classes (VOC) Challenge Classi?cation \r
Competition and the Thermal Imagery Visual Object Tracking (VOT-TIR) \r
Competition.
歡迎各位老師和同學參加。
