A Study on the Construction Method of Artificial Intelligence Training Data to Classify and Detect for Hazardous Chemicals 


Vol. 49,  No. 10, pp. 1436-1446, Oct.  2024
10.7840/kics.2024.49.10.1436


PDF
  Abstract

Recently, as the chemical industry develops, various chemical accidents are increasing. In order to respond to the increasing number of chemical accidents, various studies are being conducted on hazardous chemical accident response technology that incorporates artificial intelligence technology to detect chemical substances. Video and image-based artificial intelligence hazardous chemical detection systems require a sufficient amount of training data to enable accurate detecting of hazardous chemicals in accidents. However, due to the risk of hazardous chemicals, it is difficult to construct data, so currently, training data for artificial intelligence research for detecting hazardous chemicals is very insufficient. Therefore, this paper proposes a method of constructing an artificial intelligence training dataset that reflects the characteristics of hazardous chemicals and chemical accidents, such as the state of the material and the presence or absence of visual features. Following the proposed training dataset construction method, about 200,000 training datasets for 9 types of hazardous chemicals were constructed through collecting and securing raw data through self-experimentation of hazardous chemical , data processing, and annotation of the training data. The constructed dataset was divided in a ratio of 8:1:1 for training and validation and used training, validation, and test data. As a result of CNN-based hazardous chemical detecting, an average hazardous chemical detecting accuracy of about 90% was obtained. The detecting results are expected to provide an estimate of the accidental hazardous chemicals at the actual chemical accidents, thereby supporting appropriate and rapid response to the accidental substances to on-site firefighter.

  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]

Y. Kim, K. Choi, G. Kim, "A Study on the Construction Method of Artificial Intelligence Training Data to Classify and Detect for Hazardous Chemicals," The Journal of Korean Institute of Communications and Information Sciences, vol. 49, no. 10, pp. 1436-1446, 2024. DOI: 10.7840/kics.2024.49.10.1436.

[ACM Style]

Yeon-Jin Kim, Kap-Yong Choi, and Gyeong-Bae Kim. 2024. A Study on the Construction Method of Artificial Intelligence Training Data to Classify and Detect for Hazardous Chemicals. The Journal of Korean Institute of Communications and Information Sciences, 49, 10, (2024), 1436-1446. DOI: 10.7840/kics.2024.49.10.1436.

[KICS Style]

Yeon-Jin Kim, Kap-Yong Choi, Gyeong-Bae Kim, "A Study on the Construction Method of Artificial Intelligence Training Data to Classify and Detect for Hazardous Chemicals," The Journal of Korean Institute of Communications and Information Sciences, vol. 49, no. 10, pp. 1436-1446, 10. 2024. (https://doi.org/10.7840/kics.2024.49.10.1436)
Vol. 49, 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