Robot Eye: Automatic Object Detection and Recognition Using Deep Attention Network to Assist Blind People

Ervin Yohannes, Paul Lin, Chih Yang Lin, Timothy K. Shih

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

Abstract

Detection and Recognition is a well-known topic in computer vision that still faces many unresolved issues. One of the main contributions of this research is a method to guide blind people around an outdoor environment with the assistance of a ZED stereo camera, a camera that can calculate depth information. In this paper, we propose a deep attention network to automatically detect and recognize objects. The objects are not only limited to general people or cars, but include convenience stores and traffic lights as well, in order to help blind people cross a road and make purchases in a store. Since public datasets are limited, we also create a novel dataset with images captured by the ZED stereo camera and collected from Google Street View. When testing with images of different resolutions, our method achieves an accuracy rate of about 81%, which is better than naive YOLO v3.

Original languageEnglish
Title of host publicationProceedings - 2020 International Conference on Pervasive Artificial Intelligence, ICPAI 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages152-157
Number of pages6
ISBN (Electronic)9781665404839
DOIs
StatePublished - Dec 2020
Event1st International Conference on Pervasive Artificial Intelligence, ICPAI 2020 - Taipei, Taiwan
Duration: 3 Dec 20205 Dec 2020

Publication series

NameProceedings - 2020 International Conference on Pervasive Artificial Intelligence, ICPAI 2020

Conference

Conference1st International Conference on Pervasive Artificial Intelligence, ICPAI 2020
Country/TerritoryTaiwan
CityTaipei
Period3/12/205/12/20

Keywords

  • attention
  • deep learning
  • detection
  • navigation tools
  • recognition

Fingerprint

Dive into the research topics of 'Robot Eye: Automatic Object Detection and Recognition Using Deep Attention Network to Assist Blind People'. Together they form a unique fingerprint.

Cite this