青年博士论坛预告2017年11月15日(周三) 8:30-10:00

时间：11月15日(周三) 8:30-10:00

地点：仙林校区计算机学科楼327会议室

报告人:苗圣法

题目：时间序列的预定义模式检测

摘要:

在分析时间序列前，我们已经积累了一些关于所采集数据的先验知识。如何充分地使用先验知识来提高模式检测的效率和精确度，我提出了一种新颖的预定义模式检测方法。该预定义模式检测方法能够从时间序列中高效地检测出和预定义模式相匹配的实例，即使实例存在时间上弯曲或振幅上的形变。该方法借助模版（预定义模式）和地标（重要点）对时间序列进行压缩表示，并结合地标约束和可信区间来模拟和检测时间序列中的实例。该方法还引入了最小描述长度，对时间序列进行预处理。最小描述长度不但有助于保留时间序列中的有用信息，而且还可以防止过度拟合。

报告人简介:

苗圣法，荷兰莱顿大学高级bd手机版官网登录ios 、特温特大学行为管理和社会科学学院博后研究员，荷兰超图信息科技公司首席数据分析师。博士毕业于莱顿大学和兰州大学，获得数据挖掘专业双博士学位。硕士和本科就读于兰州大学信息科学与工程学。研究方向：数据挖掘，机器学习，大数据分析，结构质量检测，文本挖掘，风险预测。

报告人:张伟

题目：Deep learning based automatic colorization and color sketch generation

摘要:

As human intelligence is capable of analyzing and understanding images, developing automatic image transform algorithms that fulfill humans' preference, such as color and style, is important in the research of computer graphics and artificial intelligence. This talk focuses on two automatic image transform tasks, one is colorization which aims to convert grayscale images to colorful ones; another one is color sketch style transfer which is to generate color sketches from natural photographs.

We formulate image colorization as a pixel-wise prediction problem utilizing deep fully convolutional neural networks. In order to maintain color consistency in homogeneous regions as well as precisely distinguishing colors near region boundaries, we propose a novel fully automatic colorization pipeline which involves a boundary-guided CRF and a CNN-based color transform as post-processing steps. We further introduce two novel automatic evaluation schemes to efficiently assess colorization quality in terms of spatial coherence and localization. Comprehensive experiments demonstrate great quality improvement in results of our proposed colorization method under multiple evaluation metrics.

For color sketch generation, we develop a fully automatic image transform system which extends the state-of-the-art real-time neural-style transfer method. Due to the potential mismatch of the overall texture statistics between input images and the style target image, it is extremely difficult to generate color sketches using existing deep neural-style transfer method. We address this problem by selecting suitable style target online from a color sketch example set which consists 10 example color sketches. Experimental results demonstrate that our system greatly reduces artifacts compared to previous state-of-the-art methods, producing vivid color sketch results.

报告人简介:

张伟博士毕业于香港大学计算机系，此前于2013年在华中科技大学获得硕士学位，于2010年在重庆大学获得本科学位。研究方向包括深度学习，图像处理与计算机视觉等。