Efficient Training Methodology in an Image Classification Network 


Vol. 46,  No. 6, pp. 1087-1096, Jun.  2021
10.7840/kics.2021.46.6.1087


PDF
  Abstract

The recent advance of image recognition technology comes from the accumulation of numerous data and deepening of neural network. However, training these various data on a deep neural network causes various problems. Overfitting caused by a small amount of data, class imbalance resulting from the difference in the amount of data between classes, and multi-class training problems. This paper found and analyzed these problems occurring in such small data sets, and suggested solutions and analyzed the performance through experiments. For these goals, we compared open small data sets and the differences between them and selected the training techniques that perform well for each dataset.

  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.


  Cite this article

[IEEE Style]

E. Bae and S. Lee, "Efficient Training Methodology in an Image Classification Network," The Journal of Korean Institute of Communications and Information Sciences, vol. 46, no. 6, pp. 1087-1096, 2021. DOI: 10.7840/kics.2021.46.6.1087.

[ACM Style]

Eunjee Bae and Sungjin Lee. 2021. Efficient Training Methodology in an Image Classification Network. The Journal of Korean Institute of Communications and Information Sciences, 46, 6, (2021), 1087-1096. DOI: 10.7840/kics.2021.46.6.1087.

[KICS Style]

Eunjee Bae and Sungjin Lee, "Efficient Training Methodology in an Image Classification Network," The Journal of Korean Institute of Communications and Information Sciences, vol. 46, no. 6, pp. 1087-1096, 6. 2021. (https://doi.org/10.7840/kics.2021.46.6.1087)

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880