Related Experiment Video
Updated: Jan 8, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
2.2K
Crack detection in structural images using a hybrid Swin Transformer and enhanced features representation block
1Department of IoT, School of Computer Science and Engineering, Vellore Institute of Technology, Vellore, India.
Frontiers in Artificial Intelligence
|December 18, 2025
Summary
This study introduces a novel hybrid model for accurate crack detection in images, achieving 98% accuracy. The framework enhances structural monitoring by effectively identifying cracks using advanced deep learning techniques.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Structural Engineering
Background:
- Accurate crack detection is crucial for structural health monitoring.
- Traditional methods often struggle with complex image data and subtle crack patterns.
Purpose of the Study:
- To develop a robust and precise crack detection framework.
- To improve the accuracy and efficiency of identifying cracks in images for structural monitoring.
Main Methods:
- A hybrid deep learning model integrating Swin Transformer and Enhanced Features Representation Block (EFRB).
- Utilized depthwise and pointwise convolutions within EFRB for improved spatial and channel representation.
- Employed residual connections to facilitate training of deeper networks.
Main Results:
- Achieved 98% accuracy in crack detection.
- Demonstrated high performance with precision (0.97), recall (0.99), and F1-score (0.98).
- Population-based feature selection optimized the training process for robust performance.
Conclusions:
- The proposed hybrid model is highly effective for real-world crack detection applications.
- The integration of Swin Transformer and EFRB offers significant advantages in image-based structural monitoring.
- The framework provides a reliable solution for assessing structural integrity through precise crack identification.
