Weld Defect Segmentation in X-ray Image with Boundary Label Smoothing
Weld defect segmentation (WDS) is widely used to detect defects from X-ray images for welds, which is of practical importance for manufacturing in all industries. The key challenge of WDS is that the labeled ground truth of defects is usually not accurate because of the similarities between the cand...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-12-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/12/24/12818 |
| _version_ | 1797461605532827648 |
|---|---|
| author | Junhua Zhang Minghao Guo Pengzhi Chu Yang Liu Jun Chen Huanxi Liu |
| author_facet | Junhua Zhang Minghao Guo Pengzhi Chu Yang Liu Jun Chen Huanxi Liu |
| author_sort | Junhua Zhang |
| collection | DOAJ |
| description | Weld defect segmentation (WDS) is widely used to detect defects from X-ray images for welds, which is of practical importance for manufacturing in all industries. The key challenge of WDS is that the labeled ground truth of defects is usually not accurate because of the similarities between the candidate defect and noisy background, making it difficult to distinguish some critical defects, such as cracks, from the weld line during the inference stage. In this paper, we propose boundary label smoothing (BLS), which uses Gaussian Blur to soften the labels near object boundaries to provide an appropriate representation of inaccuracy and uncertainty in ground truth labels. We incorporate BLS into dice loss, in combination with focal loss and weighted cross-entropy loss as a hybrid loss, to achieve improved performance on different types of segmentation datasets. |
| first_indexed | 2024-03-09T17:21:48Z |
| format | Article |
| id | doaj.art-186b065151b24eafac11e39487b6d62c |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-09T17:21:48Z |
| publishDate | 2022-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-186b065151b24eafac11e39487b6d62c2023-11-24T13:05:18ZengMDPI AGApplied Sciences2076-34172022-12-0112241281810.3390/app122412818Weld Defect Segmentation in X-ray Image with Boundary Label SmoothingJunhua Zhang0Minghao Guo1Pengzhi Chu2Yang Liu3Jun Chen4Huanxi Liu5Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, ChinaDepartment of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, ChinaDepartment of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, ChinaDepartment of Dermatology and Department of Laser and Aesthetic Medicine, Shanghai Ninth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai 200011, ChinaDepartment of Dermatology and Dermatologic Surgery, Shanghai Ninth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai 200011, ChinaDepartment of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, ChinaWeld defect segmentation (WDS) is widely used to detect defects from X-ray images for welds, which is of practical importance for manufacturing in all industries. The key challenge of WDS is that the labeled ground truth of defects is usually not accurate because of the similarities between the candidate defect and noisy background, making it difficult to distinguish some critical defects, such as cracks, from the weld line during the inference stage. In this paper, we propose boundary label smoothing (BLS), which uses Gaussian Blur to soften the labels near object boundaries to provide an appropriate representation of inaccuracy and uncertainty in ground truth labels. We incorporate BLS into dice loss, in combination with focal loss and weighted cross-entropy loss as a hybrid loss, to achieve improved performance on different types of segmentation datasets.https://www.mdpi.com/2076-3417/12/24/12818weld defect segmentationboundary label smoothinghybrid loss |
| spellingShingle | Junhua Zhang Minghao Guo Pengzhi Chu Yang Liu Jun Chen Huanxi Liu Weld Defect Segmentation in X-ray Image with Boundary Label Smoothing Applied Sciences weld defect segmentation boundary label smoothing hybrid loss |
| title | Weld Defect Segmentation in X-ray Image with Boundary Label Smoothing |
| title_full | Weld Defect Segmentation in X-ray Image with Boundary Label Smoothing |
| title_fullStr | Weld Defect Segmentation in X-ray Image with Boundary Label Smoothing |
| title_full_unstemmed | Weld Defect Segmentation in X-ray Image with Boundary Label Smoothing |
| title_short | Weld Defect Segmentation in X-ray Image with Boundary Label Smoothing |
| title_sort | weld defect segmentation in x ray image with boundary label smoothing |
| topic | weld defect segmentation boundary label smoothing hybrid loss |
| url | https://www.mdpi.com/2076-3417/12/24/12818 |
| work_keys_str_mv | AT junhuazhang welddefectsegmentationinxrayimagewithboundarylabelsmoothing AT minghaoguo welddefectsegmentationinxrayimagewithboundarylabelsmoothing AT pengzhichu welddefectsegmentationinxrayimagewithboundarylabelsmoothing AT yangliu welddefectsegmentationinxrayimagewithboundarylabelsmoothing AT junchen welddefectsegmentationinxrayimagewithboundarylabelsmoothing AT huanxiliu welddefectsegmentationinxrayimagewithboundarylabelsmoothing |