Real-Time Deep Model for Intraoral Scanner Image Segmentation 


Vol. 48,  No. 10, pp. 1261-1270, Oct.  2023
10.7840/kics.2023.48.10.1261


PDF
  Abstract

Recently, the use of intraoral scanners for creating three-dimensional models of teeth has been increasing in dental practices. These scanners reconstruct 3D models using multiple 2D images obtained through scanning. In this process, segmentation is used to remove unnecessary objects such as fingers and tongues from the reconstruction. Real-time visualization is crucial for achieving the accuracy of the model construction; thus, the time given to the segmentation process is limited. Previous studies on real-time segmentation have achieved real-time requirements by reducing the number of convolutions using a branch structure while minimizing the decrease in accuracy. In this study, we propose the IntraOralScanner Network (IOSNet), a model that reduces branch structure dependency to decrease inference time. We improved the remaining branches' layers to enhance speed and maintain accuracy, achieving a 62.2% increase in inference speed (compared to NVIDIA GTX 1060) and only a 2.7% decrease in accuracy, as measured by mIoU, compared to the existing real-time model PIDNet-small. We achieved the segmentation time goal required for commercial scanner product release, i.e., 10ms (on the same GPU) and 30 FPS.

  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]

H. Yoo, H. Cho, J. Kim, K. Jun, "Real-Time Deep Model for Intraoral Scanner Image Segmentation," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 10, pp. 1261-1270, 2023. DOI: 10.7840/kics.2023.48.10.1261.

[ACM Style]

Hojin Yoo, Hangjae Cho, Jaegon Kim, and Kyungkoo Jun. 2023. Real-Time Deep Model for Intraoral Scanner Image Segmentation. The Journal of Korean Institute of Communications and Information Sciences, 48, 10, (2023), 1261-1270. DOI: 10.7840/kics.2023.48.10.1261.

[KICS Style]

Hojin Yoo, Hangjae Cho, Jaegon Kim, Kyungkoo Jun, "Real-Time Deep Model for Intraoral Scanner Image Segmentation," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 10, pp. 1261-1270, 10. 2023. (https://doi.org/10.7840/kics.2023.48.10.1261)
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