Recognition Effects of Deep Convolutional Neural Network on Smudged Handwritten Digits

Zhe Xu, Yusuke Terada, Dongbao Jia, Zonghui Cai, Shangce Gao

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

Deep convolutional neural network (CNN) is known to be the first truly successful deep learning approach for image processing and understanding, e.g., the handwritten digits discrimination. However, in real applications such as handwritten zip code recognition, the collected images are commonly with smudged background. In this paper, we study the recognition effects of CNN on smudged digits and compared the results with three-layered perceptron. Experimental results based on MNIST dataset with smudged background (simulated by salt-and-pepper and gaussian noises) show that a drastic decline of recognition accuracy is observed for CNN, suggesting that the extracted features by convolutional operation and max pooling is very sensitive to the noise.

Original languageEnglish
Title of host publicationProceedings - 2018 5th International Conference on Information Science and Control Engineering, ICISCE 2018
EditorsShaozi Li, Ying Dai, Yun Cheng
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages412-416
Number of pages5
ISBN (Electronic)9781538655009
DOIs
StatePublished - 2018/07/02
Event5th International Conference on Information Science and Control Engineering, ICISCE 2018 - Zhengzhou, Henan, China
Duration: 2018/07/202018/07/22

Publication series

NameProceedings - 2018 5th International Conference on Information Science and Control Engineering, ICISCE 2018

Conference

Conference5th International Conference on Information Science and Control Engineering, ICISCE 2018
Country/TerritoryChina
CityZhengzhou, Henan
Period2018/07/202018/07/22

Keywords

  • Deep convolutional neural network
  • MNIST dataset
  • handwritten digits discrimination
  • noise

ASJC Scopus subject areas

  • Decision Sciences (miscellaneous)
  • Information Systems and Management
  • Control and Systems Engineering
  • Industrial and Manufacturing Engineering

Fingerprint

Dive into the research topics of 'Recognition Effects of Deep Convolutional Neural Network on Smudged Handwritten Digits'. Together they form a unique fingerprint.

Cite this