Related Experiment Video
Updated: Apr 24, 2026
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
1.3K
A deep learning-based approach for detecting anomalous behavior in safety-critical spaces
Aqib Anees1, Syed Asim Jalal1, Hassan Jalil Hadi2
1Department of Computer Science, University of Peshawar, Peshawar, Pakistan.
Frontiers in Artificial Intelligence
|April 16, 2026
Summary
This study introduces an AI system using deep learning to detect wrong-turn traffic violations at roundabouts. The artificial intelligence (AI) model, trained on a custom dataset, shows promise for real-world traffic safety applications.
Area of Science:
- Computer Science
- Traffic Engineering
- Artificial Intelligence
Background:
- Wrong-turn violations at roundabouts are understudied despite contributing to congestion and crashes.
- A lack of dedicated datasets hinders research in detecting these specific traffic violations.
Purpose of the Study:
- To address the gap in wrong-turn violation detection by developing a deep learning approach.
- To create a novel dataset for training and evaluating wrong-turn violation detection models.
Main Methods:
- A deep learning system utilizing the You Only Look Once (YOLO) algorithm was developed.
- Video data from roundabouts was collected and a custom dataset was annotated.
- The YOLO model was trained and evaluated using accuracy and recall metrics.
Main Results:
- The developed AI model demonstrated effectiveness in detecting wrong-turn violations in real-time.
- The custom dataset provided a valuable resource for training the detection model.
Conclusions:
- The proposed deep learning approach shows significant potential for real-world implementation in traffic management.
- Further refinement of the system could enhance traffic safety at roundabouts.