HKAN: A Hybrid Kolmogorov–Arnold Network for Robust Fabric Defect Segmentation
Author Information
Author(s): Li Min, Ye Pei, Cui Shuqin, Zhu Ping, Liu Junping
Primary Institution: School of Computer and Artificial Intelligence, Wuhan Textile University, Wuhan, China
Hypothesis
Can combining CNNs with Transformers and Kolmogorov–Arnold Networks improve fabric defect segmentation?
Conclusion
The HKAN model outperforms existing semantic segmentation models in fabric defect detection.
Supporting Evidence
- HKAN achieved 99.33% pixel accuracy and 91.60% mean intersection over union on the Four-Fabric-Defects Dataset.
- HKAN outperformed all other models in terms of accuracy and computational efficiency.
- Extensive experiments demonstrated HKAN's superior performance across diverse fabric images.
Takeaway
This study created a new way to find fabric defects using a mix of different computer vision techniques, making it better at spotting problems in fabrics.
Methodology
The study developed a Hybrid KAN model that integrates CNNs, Transformers, and KANs for fabric defect segmentation, using extensive experiments on three datasets.
Limitations
The model's computational complexity may hinder its application in resource-constrained environments.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website