Scholarlink

<p>Fabric defect prediction is a difficult task in the fabric business due to its complicated shapes and wide range of fabric flaws like drop stitches, misprinting, crease marks, barre, neps/knots, and splicing etc. It is essential for quality control and fabric inspection. Models based on convolutional neural networks (CNNs) have shown efficacy in the detection of fabric defects. However, because of the sophisticated background texture of fabric, these models are suffering with limitations like more processing time and less accuracy. To address these limitations, we proposed an improved Region-based Convolutional Neural Network (R-CNN) combined with Gabor filter and Selective Search algorithm. By taking the advantage of Gabor filter in frequency analysis, and Selective search, we have improved the performance of R-CNN. We also adjusted the R-CNN pooling layer in our suggested model to enable efficient processing of feature maps, which has a major benefit in lowering the unfavorable effects of the R-CNN structure and raising prediction accuracy. The performance of enhanced R- CNN is verified on visual studio IDE 1.88 platform, and it is found that proposed model takes the less time and shows the significant improvement in accuracy. Our model achieved an impressive accuracy of 98.21%, outperforming all other models like Mobile Net, U-Net, LeNet, DenseNet, IM-RCNN, and R-CNN. This high level of accuracy demonstrates the efficacy of our approach for detecting fabric flaws.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Selective Search Based Gabor Wavelet for Fabric Defect Prediction Using Enhanced R-CNN

  • 期刊论文

  • 发表日期：

    2025年3月26日

查看全文
  • Yakobu Dasari,
  • Vijay Krishna Chebrolu,
  • M. Harsha Vardhan,
  • Saisreeja Kondru,
  • Venkata Rajulu Pilli

摘要

Fabric defect prediction is a difficult task in the fabric business due to its complicated shapes and wide range of fabric flaws like drop stitches, misprinting, crease marks, barre, neps/knots, and splicing etc. It is essential for quality control and fabric inspection. Models based on convolutional neural networks (CNNs) have shown efficacy in the detection of fabric defects. However, because of the sophisticated background texture of fabric, these models are suffering with limitations like more processing time and less accuracy. To address these limitations, we proposed an improved Region-based Convolutional Neural Network (R-CNN) combined with Gabor filter and Selective Search algorithm. By taking the advantage of Gabor filter in frequency analysis, and Selective search, we have improved the performance of R-CNN. We also adjusted the R-CNN pooling layer in our suggested model to enable efficient processing of feature maps, which has a major benefit in lowering the unfavorable effects of the R-CNN structure and raising prediction accuracy. The performance of enhanced R- CNN is verified on visual studio IDE 1.88 platform, and it is found that proposed model takes the less time and shows the significant improvement in accuracy. Our model achieved an impressive accuracy of 98.21%, outperforming all other models like Mobile Net, U-Net, LeNet, DenseNet, IM-RCNN, and R-CNN. This high level of accuracy demonstrates the efficacy of our approach for detecting fabric flaws.