Road object detection is crucial for enhancing the safety of Advanced Driver Assistance Systems (ADAS). This paper presents a model fine-tuned with YOLOv8 and the BDD dataset, specifically designed for road object detection using monocular camera input from moving vehicles. The model leverages the YOLOv8 architecture and transfer learning to accurately identify road objects in video streams captured by these vehicles. Additionally, it employs genetic evolution and mutation techniques for the systematic optimization of hyperparameters governing the YOLOv8 model. The integration of this hyperparameter evolution technique underscores the importance of optimizing model parameters to achieve superior performance for ADAS, aligning with recent advancements in the field. Through extensive experimentation, our model demonstrates significant efficacy in real-time road object detection, achieving a mean Average Precision (mAP) of 0.629 and an impressive F1 score of 0.95. Furthermore, the model processes video data at 56 frames per second (FPS), highlighting its suitability for real-time ADAS applications.