The detection and monitoring of objects in traffic environments are key factors in developing intelligent, safe, and reliable autonomous vehicle systems. However, most modern detection models, such as YOLOv8 or R-CNN, are primarily trained and evaluated under ideal environmental conditions, which do not fully reflect the real-world challenges of rain, fog, or low lighting - factors that significantly impact the accuracy of detecting pedestrians, vehicles, traffic signs, and signals. This study proposes a comprehensive experimental approach to evaluate the performance of the state-of-the-art object detection model YOLOv11 under adverse weather conditions using the BDD100K dataset. The experimental results provide valuable insights into the adaptability, stability, and processing speed of the model, informing technical recommendations for real-world deployment in real-time computer vision systems. Research contributes to bridging the gap between AI models in ideal settings and their practical applications, particularly in complex traffic scenarios and high-reliability demands such as autonomous driving and intelligent urban surveillance.