Annotated Dataset for Detecting Wrong-Turn Violations at Urban Roundabouts Using Deep Learning

This dataset contains annotated images for detecting wrong-turn traffic violations at urban roundabouts. The data were collected from multiple roundabouts in Peshawar, Pakistan, using strategically positioned roadside cameras. The dataset was developed to support research on automated traffic violation detection in safety-critical traffic environments.
The images capture different traffic scenarios, including vehicles performing correct maneuvers as well as wrong-turn violations. Each image is manually annotated with bounding boxes corresponding to relevant object classes, including cars, bikes/rickshaws, and violation instances.
The dataset was created to train and evaluate deep learning models for real-time traffic monitoring and anomaly detection. It was used in the study titled “A deep learning-based approach for detecting anomalous behavior in safety-critical spaces.”