Deep learning and its application to image classification
Abstract
Deep learning has become a new area of machine learning research based on the algorithms which attempt to extract multi-level representation (abstraction) for data such as images, audio and text. The techniques of deep learning have significant impact on a wide range of signal and information processing applications. As a foundational issue in computer vision area, image classification has drawn much attention from many researchers, as well as in industry, with for instance, Google, MSRA and Baidu, all being actively involved in deep learning research and related competition.
In this talk, the related techniques in deep learning will be first introduced. Then the recent work and progress in image classification by the speakers will be introduced in technical detail, by which the class relatedness modeling is integrated with deep representation learning, to significantly improve the classification performance by transferring knowledge appropriately.
SpeakerLi SHEN | |
Date & Time19 May 2015 (Tuesday) 15:30 - 17:00 | |
VenueE11-4045 (University of Macau) | |
Organized byDepartment of Computer and Information Science |
Biography
Li SHEN University of Chinese Academy of Sciences
Li Shen received her B.S. degree from Nankai University, Tianjin in 2009. In her university study period, she received various studentships and awards including the first rank scholarship in 2006 and the Tianjin People’s Government Scholarship award in 2008 (top 3%). She is currently a PhD candidate at University of Chinese Academy of Sciences. Currently, she mainly works on computer vision and machine learning, with particular interest in the area of object recognition, deep learning, hierarchical model and transfer learning. She has a research internship in the Internet Media Group in Microsoft Research Asia, from December 2014 to March 2015. She has published several works on top international conferences of computer vision and artificial intelligence areas, including Computer Vision and Pattern Recognition (CVPR) and International Joint Conference on Artificial Intelligence (IJCAI). The recent work has achieved state-of-the-art results on one of the challenging computer vision dataset, CIFAR-100.
Gang SUN Institute of Software, Chinese Academy of Sciences
Gang Sun received his B.S. degree from Nankai University in 2009, and in his university study period, he received various studentships and awards including the first rank scholarship, the National Scholarship award, the National Scholarship for Encouragement award andthe Excellent Graduation Dissertation award, and Outstanding Graduates award. He is currently a PhD candidate at Institute of Software, Chinese Academy of Sciences. His research interests include computer vision and deep learning, especially in image classification and convolutional neural networks. He has a research internship in Visual Computing Group in Microsoft Research Asia, and also an internship in Institute of Deep Learning (IDL) in Baidu Research. On one of the most challenging computer vision benchmarks, the ImageNet classification challenge, he achieves the best result to date, with a top-5 error rate of 4.58% and exceeding the human recognition performance, a relative 31% improvement over the ILSVRC 2014 winner.