Generative adversarial network-based data augmentation for foreign object detection with small samples in railway catenary systems

Key Points

• Foreign object detection accuracy increased with the use of data augmentation techniques.
• The model achieved a detection rate of 85% under limited sample sizes, showing its effectiveness.
• Assessment using generative adversarial networks for data augmentation enhances detection capabilities.
• Findings imply potential improvements in safety measures for railway catenary systems.