Publication detail

Cooperative Coevolutionary Approximation in HOG-based Human Detection Embedded System

WIGLASZ, M. SEKANINA, L.

Original Title

Cooperative Coevolutionary Approximation in HOG-based Human Detection Embedded System

Type

conference paper

Language

English

Original Abstract

The histogram of oriented gradients (HOG) feature extraction is a computer vision method widely used in embedded systems for detection of objects such as pedestrians. We used cooperative coevolutionary Cartesian genetic programming (CGP) to exploit the error resilience in the HOG algorithm. We evolved new approximate implementations of the arctan and square root functions, which are typically employed to compute the gradient orientations and magnitudes. When the best evolved approximations are integrated into the software implementation of the HOG algorithm, not only the execution time, but also the classification accuracy was improved in comparison with approximations evolved separately using CGP and also compared to the state-of-the art approximate implementations. As the evolved code does not contain any loops and branches, it is suitable for the follow-up low-power hardware implementation.

Keywords

Approximate computing, Cartesian genetic programming, Cooperative coevolution, Histogram of oriented gradients

Authors

WIGLASZ, M.; SEKANINA, L.

Released

8. 12. 2018

Publisher

Institute of Electrical and Electronics Engineers

Location

Bengaluru

ISBN

978-1-5386-9276-9

Book

2018 IEEE Symposium Series on Computational Intelligence (SSCI 2018)

Pages from

1313

Pages to

1320

Pages count

8

URL

https://www.fit.vut.cz/research/publication/11695/

BibTex

@inproceedings{BUT155023,
  author="Michal {Wiglasz} and Lukáš {Sekanina}",
  title="Cooperative Coevolutionary Approximation in HOG-based Human Detection Embedded System",
  booktitle="2018 IEEE Symposium Series on Computational Intelligence (SSCI 2018)",
  year="2018",
  pages="1313--1320",
  publisher="Institute of Electrical and Electronics Engineers",
  address="Bengaluru",
  doi="10.1109/SSCI.2018.8628910",
  isbn="978-1-5386-9276-9",
  url="https://www.fit.vut.cz/research/publication/11695/"
}

Documents

SS-1146.pdf