THE 3 rd INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
172
At test time we multiply the conditional class probabilities and the
individual box confidence predictions, which gives us class-specific
confidence scores for each box:
[2]
Results of the Simulation Firstly, the main steps are represented in the figure 1 below:
Figure 11. Steps
In the simulation, object detection is done by using trained YOLOv3 on
custom dataset. After preparing the custom dataset in the YOLOv3 format,
YOLOv3 is trained and tested, predictions are displayed. Simulation outputs
for the different input images are displayed below:
Figure 12. Simulation results according to input images
As it can be seen from figure 2, predictions are 100% accurate. After
the ID verification, the person’s vaccine passport and current COVID status
should be checked according to his unique id in the database.
Conclusion The accuracy and performance of real-time face recognition using the
YOLOv3 model has been evaluated multiple times. The model performed as
planned and can be further developed depending on the application field.
Moreover, it is better to use YOLOv3 for only extracting faces because it
requires a lot of images per person for training. Therefore, obtained image of
face by YOLOv3 is sent to CNN in order to identify the person. CNN model
would require less images per person compared to YOLOv3.
References [1] Real-Time Object Detection using YOLO: A review, Upulie Handalage, Lakshini
Kuganandamurthy, 2021, 3-4
[2] You Only Look Once: Unified, Real-Time Object Detection Joseph Redmon, Santosh
Divvala, Ross Girshick, Ali Farhadi, 2016, 1-2