Chinese Journal of Liquid Crystals and Displays, Volume. 39, Issue 1, 59(2024)
X-ray weld defect detection method based on dense connection and multi-scale pooling
Fig. 8. Segmentation results before and after adding attention module
Fig. 10. Segmentation results of this method on the dataset provided by a non-destructive testing company
Table 1. Ablation experiment results of adding different module ablation
View tableView in ArticleTable 1. Ablation experiment results of adding different module ablation
模型 Dice/% VOE/% RVD/% MSSD/mm Unet 87.24 28.89 -12.41 17.73 Unet+密集连接空洞卷积 90.55 26.42 -11.57 14.62 Unet+多卷积核池化 90.88 25.66 -10.02 13.78 Unet+DP_block 92.33 23.35 -8.44 12.17 DP_Unet（Unet+DP_block+GAM注意力机制） 93.45 23.09 -7.01 11.32
Table 2. Quantitative results of different defect segmentation methods
View tableView in ArticleTable 2. Quantitative results of different defect segmentation methods
方法 Dice/% VOE/% RVD/% MSSD/mm Unet 87.24 28.89 -12.41 17.73 U2net 86.11 28.66 -11.86 16.99 Linknet 90.57 26.41 -9.47 13.67 DeepLabv3+ 91.66 24.67 -8.68 11.07 TransUnet 92.07 24.32 -8.35 12.44 本文方法 93.45 23.09 -7.01 11.32
Table 3. Quantitative analysis of this method on data sets provided by non-destructive testing companies
View tableView in ArticleTable 3. Quantitative analysis of this method on data sets provided by non-destructive testing companies
Dice/% VOE/% RVD/% MSSD/mm 90.36 23.97 -9.06 12.87
Get Citation
Copy Citation Text
Yong ZHANG, Peng WANG, Zhigang LÜ, Ruohai DI, Xiaoyan LI, Liangliang LI. X-ray weld defect detection method based on dense connection and multi-scale pooling[J]. Chinese Journal of Liquid Crystals and Displays, 2024, 39(1): 59
Category: Research Articles
Received: Mar. 6, 2023
Accepted: --
Published Online: Mar. 27, 2024
The Author Email: Peng WANG (wp_xatu@163.com)
