专栏名称: AI数据派

THU数据派"基于清华，放眼世界"，以扎实的理工功底闯荡“数据江湖”。发布全球大数据资讯，定期组织线下活动，分享前沿产业动态。了解清华大数据，敬请关注姐妹号“数据派THU”。

近期活动盘点：二值化网络 & 基于素描图的图像检索、2018数据与媒介发展论坛、大数据应用中日交流论坛（11.04-11.15）

AI数据派 · 公众号 · · 2018-11-04 07:30

正文

想知道近期有什么最新活动？大数点为你整理的近期活动信息在此：

二值化网络 & 基于素描图的图像检索

2018年11月04日

【时间】 2018年11月04日（周日）下午2:00 -- 4:00

【地址】 中国科学院自动化研究所

【报名方式】 扫描下方 二维码 报名

【部分精彩内容提纲】

1.杨朝晖部分：二值化网络是将网络的参数和激活值二值化，在不改变网络结构的条件下压缩网络并降低运算时间，本次报告主要介绍二值化网络的发展，包括BinaryConnect, BNN, XNORnet, DoReFa-Net等经典模型。

2.庞璐部分：Compared to text, the important characteristic of sketches lies with their ability of capturing object structure and appearance. Therefore, as a query modality, sketches are employed on many retrieval applications. With the rapid development of touch-screen devices, fine-grained sketch-based image retrieval(FG-SBIR) has become an increasingly research topic problem recently. Free-hand sketches are used in commercial search scenarios such as searching a specific shoe or chair. Besides, in forensic applications, there are no photo of the suspect for retrieving, which have motivated research on forensic facial sketches matching and sketch re-identification. Common characteristics of these problems lie with challenges, which is cross-domain and instance-level. This report provides a comprehensive introduction of established methods about grained sketch-based image retrieval. Moreover, I will discuss about recent developments on sketch re-identification.