• 综合
  • 标题
  • 关键词
  • 摘要
  • 学者
  • 期刊-刊名
  • 期刊-ISSN
  • 会议名称
搜索

作者:

Pang, Junbiao (Pang, Junbiao.) (学者:庞俊彪) | Lin, Huihuang (Lin, Huihuang.) | Su, Li (Su, Li.) | Zhang, Chunjie (Zhang, Chunjie.) | Zhang, Weigang (Zhang, Weigang.) | Duan, Lijuan (Duan, Lijuan.) (学者:段立娟) | Huang, Qingming (Huang, Qingming.) | Yin, Baocai (Yin, Baocai.) (学者:尹宝才)

收录:

EI Scopus

摘要:

Convolutional Neural Networks (CNNs) have delivered impressive state-of-the-art performances for many vision tasks, while the computation costs of these networks during test-time are notorious. Empirical results have discovered that CNNs have learned the redundant representations both within and across different layers. When CNNs are applied for binary classification, we investigate a method to exploit this redundancy across layers, and construct a cascade of classifiers which explicitly balances classification accuracy and hierarchical feature extraction costs. Our method cost-sensitively selects feature points across several layers from trained networks and embeds non-expensive yet discriminative features into a cascade. Experiments on binary classification demonstrate that our framework leads to drastic test-time improvements, e.g., possible 47.2x speedup for TRECVID upper body detection, 2.82x speedup for Pascal VOC2007 People detection, 3.72x for INRIA Person detection with less than 0.5% drop in accuracies of the original networks. © 2016 IEEE.

关键词:

Cascades (fluid mechanics) Classification (of information) Convolution Feature extraction Image processing Network layers Neural networks

作者机构:

  • [ 1 ] [Pang, Junbiao]Beijing Key Laboratory of Multimedia and Intelligent Software Technology, China
  • [ 2 ] [Pang, Junbiao]College of Metropolitan Transportation, Beijing University of Technology, China
  • [ 3 ] [Lin, Huihuang]Beijing Key Laboratory of Multimedia and Intelligent Software Technology, China
  • [ 4 ] [Lin, Huihuang]College of Metropolitan Transportation, Beijing University of Technology, China
  • [ 5 ] [Su, Li]School of Computer and Control Engineering, University of Chinese Academy of Sciences, China
  • [ 6 ] [Su, Li]Key Lab on Big Data Mining and KnowledgeManagement, Chinese Academy of Sciences, China
  • [ 7 ] [Zhang, Chunjie]School of Computer and Control Engineering, University of Chinese Academy of Sciences, China
  • [ 8 ] [Zhang, Chunjie]Key Lab on Big Data Mining and KnowledgeManagement, Chinese Academy of Sciences, China
  • [ 9 ] [Zhang, Weigang]School of Computer Science and Technology, Harbin Institute of Technology at Weihai, China
  • [ 10 ] [Duan, Lijuan]College of Computer Science and Technology, Beijing University of Technology, China
  • [ 11 ] [Huang, Qingming]School of Computer and Control Engineering, University of Chinese Academy of Sciences, China
  • [ 12 ] [Huang, Qingming]Key Lab on Big Data Mining and KnowledgeManagement, Chinese Academy of Sciences, China
  • [ 13 ] [Yin, Baocai]Beijing Key Laboratory of Multimedia and Intelligent Software Technology, China
  • [ 14 ] [Yin, Baocai]College of Metropolitan Transportation, Beijing University of Technology, China

通讯作者信息:

电子邮件地址:

查看成果更多字段

相关关键词:

相关文章:

来源 :

ISSN: 1522-4880

年份: 2016

卷: 2016-August

页码: 1037-1041

语种: 英文

被引次数:

WoS核心集被引频次: 0

SCOPUS被引频次: 3

ESI高被引论文在榜: 0 展开所有

万方被引频次:

中文被引频次:

近30日浏览量: 2

在线人数/总访问数:1035/3595618
地址:北京工业大学图书馆(北京市朝阳区平乐园100号 邮编:100124) 联系我们:010-67392185
版权所有:北京工业大学图书馆 站点建设与维护:北京爱琴海乐之技术有限公司