Parallel U-Net Based Semantic Segmentation Method Using Generated Data from YOLO V5 


Vol. 48,  No. 3, pp. 319-326, Mar.  2023
10.7840/kics.2023.48.3.319


PDF
  Abstract

In this study, we propose and implement an sementic segmentation method using a pair of U-Net in parallel where its training data are images generated from the YOLO(You Only Look Once) V5 object detection technique using yolov5s model. Image for vehicles and pedestrians from the YOLO model are in the form of bounding box and they are used for training data of a parallel U-Net. The proposed U-Net receives training images for learning in such a way that one U-Net receives original images and the other parallelly connected U-Net receives output images from YOLO V5, and then performs sementic segmentation. We compared the detection performance of the proposed method with the conventional semantic segmentation methods using a U-Net or a U-Net3+. BDD100K(Berkeley Deep Drive 100K) data set was used as training data for YOLO V5, and we found that mIoU(Mean Intersection over Union) of the proposed method is about 0.1~0.2 higher than a conventional U-Net and 0.03~0.15 higher than U-Net3+.

  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]

Guk-HanJo, Kwang-MinHyun, Young-JoonSong, "Parallel U-Net Based Semantic Segmentation Method Using Generated Data from YOLO V5," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 3, pp. 319-326, 2023. DOI: 10.7840/kics.2023.48.3.319.

[ACM Style]

Guk-HanJo, Kwang-MinHyun, and Young-JoonSong. 2023. Parallel U-Net Based Semantic Segmentation Method Using Generated Data from YOLO V5. The Journal of Korean Institute of Communications and Information Sciences, 48, 3, (2023), 319-326. DOI: 10.7840/kics.2023.48.3.319.

[KICS Style]

Guk-HanJo, Kwang-MinHyun, Young-JoonSong, "Parallel U-Net Based Semantic Segmentation Method Using Generated Data from YOLO V5," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 3, pp. 319-326, 3. 2023. (https://doi.org/10.7840/kics.2023.48.3.319)
Vol. 48, No. 3 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