Enhancing computer vision-based bridge traffic identification at nighttime through CycleGAN-enabled data augmentation

Key Points

• Results indicate a significant improvement in traffic identification accuracy, particularly at nighttime, enhancing safety.
• The analysis highlighted a 25% increase in accuracy over the standard approach, emphasizing the efficacy of CycleGAN.
• Application of CycleGAN enabled effective data augmentation, addressing the scarcity of nighttime traffic images, resulting in better training models.
• Findings suggest that using CycleGAN for data augmentation may lead to more reliable traffic identification systems.