A REGION-SPECIFIC APPROACH TO TRAFFIC SIGNRECOGNITION IN KAZAKHSTAN: A COMPARATIVE STUDY OFRESNET-101, MOBILENETV2, AND YOLOV8
DOI:
https://doi.org/10.47344/sdubnts.v65i2.1271Keywords:
Traffic Sign Recognition, Deep Learning, YOLOv8, ResNet-101, MobileNetV2, Data Augmentation, Advanced Driver Assistance System (ADAS)Abstract
This research addresses the critical need for accurate traffic sign recognition in Kazakhstan, which is essential for enhancing road safety and
developing advanced driver-assistance systems (ADAS). We created a comprehensive dataset tailored to Kazakhstan's traffic conditions and evaluated three state-of-the-art deep learning models: ResNet-101, MobileNetV2, and
YOLOv8. Among these, YOLOv8 demonstrated superior performance, achieving 89.2% accuracy, 89.6% precision, 88.9% recall, and an 89.2% F1-
score. This study highlights the effectiveness of tailored data augmentation techniques and the potential of YOLOv8 for real-time traffic sign recognition in dynamic environments, significantly contributing to the improvement of ADAS and road safety in Kazakhstan
