Iranian traffic sign dataset: A step towards understanding smart cities

Area of Science:

• Computer Science
• Artificial Intelligence
• Computer Vision

Background:

• Smart city expansion necessitates real-time road infrastructure analysis.
• Accurate perception and recording of environmental data are crucial for intelligent systems.
• Developing AI for traffic sign detection is vital for road safety and navigation.

Purpose of the Study:

• To introduce a novel system for detecting Iranian traffic signs.
• To construct a comprehensive dataset of Iranian traffic signs.
• To develop an automated detection system using deep neural networks.

Main Methods:

• Collected a unique dataset of 14,111 images with 19,000 traffic signs across 118 classes.
• Annotated images using MakeSense tool in YOLO and Pascal VOC formats.
• Developed an automated detection system using the YOLOv12 deep neural network model.

Main Results:

• Achieved outstanding accuracy with a mean Average Precision at 50% (mAP@50) of 96% through 6-Fold Cross Validation.
• Demonstrated superior performance compared to previous YOLO versions in real-world scenarios.
• The system proved efficient in identifying diverse traffic signs under various conditions.

Conclusions:

• The developed AI system offers precise traffic sign detection, contributing to road safety.
• The system can generate digital traffic sign maps for navigation and intelligent vehicles.
• This research advances smart city infrastructure and national road safety initiatives.