Deep actionlet proposals for driver's behavior monitoring

Anwaar Ul-Haq, Jing He, Yanchun Zhang

Research output: Book chapter/Published conference paperConference paper

Abstract

Automated cell phone usage detection systems are a privilege for traffic law enforcement to avoid increasing death toll due to irresponsible drive behavior This paper presents a computer vision based approach to automatically monitor driver's behavior during driving especially focusing on irresponsible behavior like use of mobile phone. Majority of the earlier work is based on the camera system inside the vehicle. Our framework however, is based on traffic monitoring cameras on the road. At first, we learn deep action specific features based on pose and appearance representation and train an SVM, named deep actionlet. We then use action-let activations from input images to get candidate actionlet proposals. Finally, we integrate these proposals into a Faster spectral-RCNN replacing region proposal network. The spectral domain provides a significant speedup with underlying model unchanged. The proposed approach is evaluated on the challenging Strategic Highway Research Program (SHRP2) and internally developed Driver in the Wild dataset. Detailed experimental results show that our approach achieves better performance compared to the state-of-the art approaches.
Original languageEnglish
Title of host publication2017 International Conference on Image and Vision Computing New Zealand (IVCNZ)
Place of PublicationUnited States
PublisherIEEE, Institute of Electrical and Electronics Engineers
Pages1-6
Number of pages6
ISBN (Print)9781538642764
DOIs
Publication statusPublished - 2018
Event32nd International Image and Vision Computing New Zealand Conference: IVCNZ 2017 - University of Canterbury , Christchurch, New Zealand
Duration: 04 Dec 201706 Dec 2017
http://ivcnz2017.canterbury.ac.nz/ (Conference website)
https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8396321 (Conference proceedings)

Conference

Conference32nd International Image and Vision Computing New Zealand Conference
CountryNew Zealand
CityChristchurch
Period04/12/1706/12/17
OtherImage and Vision Computing New Zealand is New Zealand's premier academic conference on all aspects of computer vision, image processing, visualisation and computer graphics.
Internet address

Fingerprint Dive into the research topics of 'Deep actionlet proposals for driver's behavior monitoring'. Together they form a unique fingerprint.

  • Cite this

    Ul-Haq, A., He, J., & Zhang, Y. (2018). Deep actionlet proposals for driver's behavior monitoring. In 2017 International Conference on Image and Vision Computing New Zealand (IVCNZ) (pp. 1-6). IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/IVCNZ.2017.8402447