A Study on Rotated Animal Detection Neural Network for Image-Based Livestock Management 


Vol. 48,  No. 10, pp. 1330-1339, Oct.  2023
10.7840/kics.2023.48.10.1330


PDF
  Abstract

This paper introduces a rotated animal detection neural network for livestock management utilizing image analysis. The neural network estimates the width, length, and rotation direction of the animal's torso to accurately define the detection area of each individual livestock animal in the image. Unlike the existing method that expresses the detection area as a rectangle, our approach minimizes the surrounding area except for the animal's body, regardless of its direction in the image. This makes it applicable in livestock management environments with densely packed animals. We prepared 26,208 images of pig farms from 7 farms and 1 virtual farm to verify our proposed method. We trained our proposed method and the existing rectangular output format using this data and evaluated both using 1,333 unseen/untrained images. Our proposed method showed an mAP50 value of 0.9534, outperforming 9.04% better than the conventional method. Additionally, we confirmed that the proposed neural network has a low computational load of 13.0 GFLOPS and can be implemented at a processing rate of 29.87 FPS on an embedded device with a Maxwell GPU with 128 cores. Therefore, our proposed neural network can be effectively employed in an edge computing paradigm, even at livestock sites without internet access.

  Statistics
Cumulative Counts from November, 2022
Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


  Related Articles
  Cite this article

[IEEE Style]

K. M. Jeon, H. Ahn, S. Ju, I. Ryu, S. Kwon, J. Jung, Y. Lee, "A Study on Rotated Animal Detection Neural Network for Image-Based Livestock Management," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 10, pp. 1330-1339, 2023. DOI: 10.7840/kics.2023.48.10.1330.

[ACM Style]

Kwang Myung Jeon, HyungJun Ahn, Soheun Ju, Inchul Ryu, SolBeen Kwon, Jinwoo Jung, and Younggie Lee. 2023. A Study on Rotated Animal Detection Neural Network for Image-Based Livestock Management. The Journal of Korean Institute of Communications and Information Sciences, 48, 10, (2023), 1330-1339. DOI: 10.7840/kics.2023.48.10.1330.

[KICS Style]

Kwang Myung Jeon, HyungJun Ahn, Soheun Ju, Inchul Ryu, SolBeen Kwon, Jinwoo Jung, Younggie Lee, "A Study on Rotated Animal Detection Neural Network for Image-Based Livestock Management," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 10, pp. 1330-1339, 10. 2023. (https://doi.org/10.7840/kics.2023.48.10.1330)
Vol. 48, No. 10 Index

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880