ABSTRACT
Advanced Driver Assistance Systems (ADAS) play a vital role in enhancing road safety and minimizing human error in traffic environments. This project introduces a real-time traffic sign and road element detection system using a custom-trained YOLOv5 deep learning model. The system is capable of accurately identifying and localizing 16 distinct classes relevant to intelligent traffic control and vehicle guidance, including pedestrian crossings, road humps, traffic lights (red, yellow, green), directional signs (left, right, U-turn, no U-turn), regulatory signs (stop, no entry, give way, no horn, no parking), and pedestrian presence. The YOLOv5 model was trained on a comprehensive dataset of annotated images under diverse environmental and lighting conditions. It demonstrates high detection accuracy and real-time inference capability when tested on video inputs and webcam feeds. This implementation illustrates the potential of YOLOv5 in ADAS applications by providing timely and precise visual feedback to support autonomous navigation and decision-making, thereby contributing to safer and smarter transportation systems.

Keywords
ADAS, YOLOv5, Traffic Sign Recognition, Real-time Detection, Road Safety, Deep Learning, Object Detection, Convolutional Neural Networks, Autonomous Driving, Intelligent Transportation Systems, Pedestrian Detection, Traffic Signal Detection.

OBJECTIVE
The primary objective of this project is to design and implement a real-time traffic sign detection system for Advanced Driver Assistance Systems (ADAS) using a custom-trained YOLOv5 deep learning model. The specific goals include:
1. To develop a robust object detection model capable of accurately identifying and classifying 16 different traffic-related classes, including signals, road markings, and regulatory signs.
2. To build and preprocess a high-quality dataset of annotated traffic sign images captured under various lighting and environmental conditions.
3. To train and optimize a YOLOv5 model for real-time performance, ensuring high precision and recall in both static images and dynamic video input.
4. To implement a real-time detection pipeline that operates on live webcam or video streams and visually displays detected signs with bounding boxes and class labels.
5. To evaluate the system’s performance using standard metrics such as precision, recall, mean Average Precision (mAP), and inference speed (FPS), validating its readiness for integration into ADAS applications.