Many factors cause damage to the asphalt layer, including vehicle load, weather conditions, and the quality of the material used. Traditional investigation methods have begun to be abandoned because they are time-consuming and costly and tend to be destructive. The use of ground penetrating radar (GPR) as a non-destructive method is very often used because it is practical and more cost-effective. Therefore, the application of a machine learning model is needed to help the process of identifying GPR scan results. The objective of this research is to develop and evaluate an automated object detection model using YOLOv8 to accurately identify features such as voids, manholes, and speed bumps in asphalt layers from 3D GPR scan data. In this study, a car-mounted 3D GPR was used to scan the asphalt layer. A total of 775 object data from the 3D GPR scan results was identified with the YOLOv8 model as an automatic object detection method in the asphalt layer. The detection results show good performance with values on the metrics/precision (B) and metrics/recall (B) curves approaching 1. In addition, the average confidence values for image detection on manhole objects, speed bumps, and voids are 0.920, 0.943, and 0.924, respectively. For future research we realize that the variety of datasets used is very limited, therefore more dataset variations and real data validation are needed.