Traffic sign detection is an important function for smart transportation systems and road safety, especially within ADAS. Conventional computer vision methods and previous deep learning algorithms are poor in addressing issues like illumination variation, occlusion, and detection of smaller-sized signs. To address these limitations, this research uses YOLOv8, the new generation of the You Only Look Once family, and tailors it for traffic sign detection using dataset-specific preprocessing, fine-tuning, and ablation-driven model parameter optimization. On the Traffic Sign Dataset, YOLOv8 has a mean Average Precision (mAP) of 94.8% at 30 FPS, more than the YOLOv5 (91.9%) and SSD (88.5%) versions. Comparative testing, ablation tests, and real-world verifications ensure YOLOv8’s efficacy in delivering a better trade-off between detection performance and inference speed. The findings determine YOLOv8 to be a resilient and efficient traffic sign recognition solution that can enable safer and wiser transportation networks.