Traffic sign dataset for connected and automated vehicle operations in rural areas

Accurate traffic sign recognition is essential for the safe operation of Connected and Automated Vehicles (CAVs). However, many existing datasets, such as the LISA Traffic Sign Dataset, are predominantly composed of signs found in urban environments, offering limited representation of those common in rural settings. To address this gap, this article introduces an augmented LISA Traffic Sign Dataset that includes rural-specific traffic signs such as deer crossing, cattle crossing, and farm machinery warning signs. The dataset was compiled through a combination of real-world data collection using frames extracted from rural driving videos, and screenshots taken from google map street view. The dataset increased the number of classes in the original LISA Traffic Sign Dataset from 47 to 55 classes. Experiments conducted using a YOLOv11 -based recognition model demonstrated that the augmented dataset improved recognition performance, achieving an overall mean Average Precision (mAP) of 98.7 %, compared to 97.0 % with the original dataset. Precision and recall for the augmented dataset also increased form 95.5 % and 95.1 % to 96.7 % and 97.5 %, respectively. This expanded dataset is intended to support the development and benchmarking of recognition systems for rural CAV deployments and is structured to enable reuse in related research areas, including infrastructure assessment, rural navigation systems, and intelligent transportation studies.