CNNs(Convolution Neural Networks): the Principle and Technical Details
报告人：Aixiang Chen (Guangdong University of Finance and Economics)
时间：2018-07-23 17:00 - 2018-07-23 18:00
地点：Room 29, Quan Zhai, BICMR
摘要： CNNs赋予了计算机人眼识图的能力。本讲座通过一个简单直观的图像识别问题介绍图像处理领域里最流行的卷积神经网络的技术原理和细节.以一种直观易理解的形式解释卷积神经网络中的诸如局部感受野、参数共享、池化等核心概念及其背后特征识别的机制，最后介绍CNNs与生俱有的本质缺陷。 预期通过本讲座能让人明白CNNs如何及为何能在图像识别领域超越人眼。
Abstract: CNNs Make the computer having a capability like human eyes on recognizing object(s) in an image. This is a talk for discussing the basic principle and technical details of CNNs. In an intuitive and easily understandable way, the most of concepts such as local receptive field, parameter sharing, pooling and the mechanism of feature recognition in CNNs will be shown. Also the inherent defects in CNNs will be discussed. Through this lecture, people can understand how and why CNNs can surpass the human eye in the field of image recognition.\n<\/p>\n
