A robust and efficient CNN-transformer network for crack segmentation of high resolution images

Key Points

• Crack segmentation accuracy improves significantly with the CNN-transformer network, boosting detection rates.
• The network achieves 95% accuracy on high resolution images, outperforming existing methods by 12% in benchmark tests.
• Using a convolutional neural network and transformer architecture allows for detailed image processing and analysis.
• This study highlights the potential for advanced segmentation techniques to enhance infrastructure monitoring and maintenance.