Automatic Crack Detection Using Weakly Supervised Semantic Segmentation Network and Mixed-Label Training Strategy
Automatic crack detection in construction facilities is a challenging yet crucial task. However, existing deep learning (DL)-based semantic segmentation methods for this field are based on fully supervised learning models and pixel-level manual annotation, which are time-consuming and labor-intensiv...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Sciendo
2024-02-01
|
| Series: | Foundations of Computing and Decision Sciences |
| Subjects: | |
| Online Access: | https://doi.org/10.2478/fcds-2024-0007 |
| _version_ | 1797303102714413056 |
|---|---|
| author | Zhang Shuyuan Xu Hongli Zhu Xiaoran Xie Lipeng |
| author_facet | Zhang Shuyuan Xu Hongli Zhu Xiaoran Xie Lipeng |
| author_sort | Zhang Shuyuan |
| collection | DOAJ |
| description | Automatic crack detection in construction facilities is a challenging yet crucial task. However, existing deep learning (DL)-based semantic segmentation methods for this field are based on fully supervised learning models and pixel-level manual annotation, which are time-consuming and labor-intensive. To solve this problem, this paper proposes a novel crack semantic segmentation network using weakly supervised approach and mixed-label training strategy. Firstly, an image patch-level classifier of crack is trained to generate a coarse localization map for automatic pseudo-labeling of cracks combined with a thresholding-based method. Then, we integrated the pseudo-annotated with manual-annotated samples with a ratio of 4:1 to train the crack segmentation network with a mixed-label training strategy, in which the manual labels were assigned with a higher weight value. The experimental data on two public datasets demonstrate that our proposed method achieves a comparable accuracy with the fully supervised methods, reducing over 65% of the manual annotation workload. |
| first_indexed | 2024-03-07T23:48:07Z |
| format | Article |
| id | doaj.art-62317abc42c14c2caa3cf16131515499 |
| institution | Directory Open Access Journal |
| issn | 2300-3405 |
| language | English |
| last_indexed | 2024-03-07T23:48:07Z |
| publishDate | 2024-02-01 |
| publisher | Sciendo |
| record_format | Article |
| series | Foundations of Computing and Decision Sciences |
| spelling | doaj.art-62317abc42c14c2caa3cf161315154992024-02-19T09:03:41ZengSciendoFoundations of Computing and Decision Sciences2300-34052024-02-014919511810.2478/fcds-2024-0007Automatic Crack Detection Using Weakly Supervised Semantic Segmentation Network and Mixed-Label Training StrategyZhang Shuyuan0Xu Hongli1Zhu Xiaoran2Xie Lipeng31School of Cyber Science and Engineering, Zhengzhou University, 450000, Zhengzhou, China1School of Cyber Science and Engineering, Zhengzhou University, 450000, Zhengzhou, China2Hanwei Electronics Group Corporation, 450000, Zhengzhou, China2Hanwei Electronics Group Corporation, 450000, Zhengzhou, ChinaAutomatic crack detection in construction facilities is a challenging yet crucial task. However, existing deep learning (DL)-based semantic segmentation methods for this field are based on fully supervised learning models and pixel-level manual annotation, which are time-consuming and labor-intensive. To solve this problem, this paper proposes a novel crack semantic segmentation network using weakly supervised approach and mixed-label training strategy. Firstly, an image patch-level classifier of crack is trained to generate a coarse localization map for automatic pseudo-labeling of cracks combined with a thresholding-based method. Then, we integrated the pseudo-annotated with manual-annotated samples with a ratio of 4:1 to train the crack segmentation network with a mixed-label training strategy, in which the manual labels were assigned with a higher weight value. The experimental data on two public datasets demonstrate that our proposed method achieves a comparable accuracy with the fully supervised methods, reducing over 65% of the manual annotation workload.https://doi.org/10.2478/fcds-2024-0007weakly supervised learningcrack detectionsemantic segmentationmixed label |
| spellingShingle | Zhang Shuyuan Xu Hongli Zhu Xiaoran Xie Lipeng Automatic Crack Detection Using Weakly Supervised Semantic Segmentation Network and Mixed-Label Training Strategy Foundations of Computing and Decision Sciences weakly supervised learning crack detection semantic segmentation mixed label |
| title | Automatic Crack Detection Using Weakly Supervised Semantic Segmentation Network and Mixed-Label Training Strategy |
| title_full | Automatic Crack Detection Using Weakly Supervised Semantic Segmentation Network and Mixed-Label Training Strategy |
| title_fullStr | Automatic Crack Detection Using Weakly Supervised Semantic Segmentation Network and Mixed-Label Training Strategy |
| title_full_unstemmed | Automatic Crack Detection Using Weakly Supervised Semantic Segmentation Network and Mixed-Label Training Strategy |
| title_short | Automatic Crack Detection Using Weakly Supervised Semantic Segmentation Network and Mixed-Label Training Strategy |
| title_sort | automatic crack detection using weakly supervised semantic segmentation network and mixed label training strategy |
| topic | weakly supervised learning crack detection semantic segmentation mixed label |
| url | https://doi.org/10.2478/fcds-2024-0007 |
| work_keys_str_mv | AT zhangshuyuan automaticcrackdetectionusingweaklysupervisedsemanticsegmentationnetworkandmixedlabeltrainingstrategy AT xuhongli automaticcrackdetectionusingweaklysupervisedsemanticsegmentationnetworkandmixedlabeltrainingstrategy AT zhuxiaoran automaticcrackdetectionusingweaklysupervisedsemanticsegmentationnetworkandmixedlabeltrainingstrategy AT xielipeng automaticcrackdetectionusingweaklysupervisedsemanticsegmentationnetworkandmixedlabeltrainingstrategy |