The Effect of Data Generated by GAN Method on Embedded Target Detection with YOLO v5 Model

EasyChair Preprint no. 5459

Abstract

The detection of buried targets is a difficult process due to various disruptive effects. As a solution to these difficulties, YOLO v5 algorithm, which has achieved successful results in object detection problems, has come to the fore. In this study,  simulated data consisting of B-scans were obtained with the gprMax program, since we did not have real data. Then, the number of data simulated with the Generative Adversarial Network (GAN) method was increased. The aim is to evaluate the usability of the data obtained by the GAN algorithm in embedded target detection and its effect on learning in the training through the YOLO v5 algorithm. In the proposed model, the performance of the trainings made with the data set consisting of different features was analyzed numerically based on the metric values ​​obtained.

Keyphrases: gprMax, Yere Nüfuz Eden Radar (YNR), YOLO v5, Üretken Çekişmeli Ağlar (ÜÇA)

@Booklet{EasyChair:5459,
  author = {Sertan AkÇali and Fatih Erden},
  title = {The Effect of Data Generated by GAN Method on Embedded Target Detection with YOLO v5 Model},
  howpublished = {EasyChair Preprint no. 5459},

  year = {EasyChair, 2021}}