Interpretability Analysis of Convolutional Neural Networks for Crack Detection
Crack detection is an important task in bridge health monitoring, and related detection methods have gradually shifted from traditional manual methods to intelligent approaches with convolutional neural networks (CNNs) in recent years. Due to the opaque process of training and operating CNNs, if the...
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-12-01
|
| Series: | Buildings |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2075-5309/13/12/3095 |
| _version_ | 1827575369742417920 |
|---|---|
| author | Jie Wu Yongjin He Chengyu Xu Xiaoping Jia Yule Huang Qianru Chen Chuyue Huang Armin Dadras Eslamlou Shiping Huang |
| author_facet | Jie Wu Yongjin He Chengyu Xu Xiaoping Jia Yule Huang Qianru Chen Chuyue Huang Armin Dadras Eslamlou Shiping Huang |
| author_sort | Jie Wu |
| collection | DOAJ |
| description | Crack detection is an important task in bridge health monitoring, and related detection methods have gradually shifted from traditional manual methods to intelligent approaches with convolutional neural networks (CNNs) in recent years. Due to the opaque process of training and operating CNNs, if the learned features for identifying cracks in the network are not evaluated, it may lead to safety risks. In this study, to evaluate the recognition basis of different crack detection networks; several crack detection CNNs are trained using the same training conditions. Afterwards, several crack images are used to construct a dataset, which are used to interpret and analyze the trained networks and obtain the learned features for identifying cracks. Additionally, a crack identification performance criterion based on interpretability analysis is proposed. Finally, a training framework is introduced based on the issues reflected in the interpretability analysis. |
| first_indexed | 2024-03-08T20:55:57Z |
| format | Article |
| id | doaj.art-a1d0825cb9eb40749b48775417fc1354 |
| institution | Directory Open Access Journal |
| issn | 2075-5309 |
| language | English |
| last_indexed | 2024-03-08T20:55:57Z |
| publishDate | 2023-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Buildings |
| spelling | doaj.art-a1d0825cb9eb40749b48775417fc13542023-12-22T13:58:28ZengMDPI AGBuildings2075-53092023-12-011312309510.3390/buildings13123095Interpretability Analysis of Convolutional Neural Networks for Crack DetectionJie Wu0Yongjin He1Chengyu Xu2Xiaoping Jia3Yule Huang4Qianru Chen5Chuyue Huang6Armin Dadras Eslamlou7Shiping Huang8School of Civil Engineering and Architecture, Wuhan Polytechnic University, Wuhan 430023, ChinaSchool of Civil Engineering and Transportation, South China University of Technology, Guangzhou 510640, ChinaChina Railway 17th Bureau Group (Guangzhou) Co., Ltd., Guangzhou 510799, ChinaChina Railway 17th Bureau Group (Guangzhou) Co., Ltd., Guangzhou 510799, ChinaSchool of Civil Engineering and Transportation, South China University of Technology, Guangzhou 510640, ChinaSchool of Electronics and Information Engineering, South China Normal University, Foshan 528225, ChinaSchool of Civil Engineering and Architecture, Wuhan Polytechnic University, Wuhan 430023, ChinaSchool of Civil Engineering and Transportation, South China University of Technology, Guangzhou 510640, ChinaSchool of Civil Engineering and Transportation, South China University of Technology, Guangzhou 510640, ChinaCrack detection is an important task in bridge health monitoring, and related detection methods have gradually shifted from traditional manual methods to intelligent approaches with convolutional neural networks (CNNs) in recent years. Due to the opaque process of training and operating CNNs, if the learned features for identifying cracks in the network are not evaluated, it may lead to safety risks. In this study, to evaluate the recognition basis of different crack detection networks; several crack detection CNNs are trained using the same training conditions. Afterwards, several crack images are used to construct a dataset, which are used to interpret and analyze the trained networks and obtain the learned features for identifying cracks. Additionally, a crack identification performance criterion based on interpretability analysis is proposed. Finally, a training framework is introduced based on the issues reflected in the interpretability analysis.https://www.mdpi.com/2075-5309/13/12/3095structural health monitoringcrack detectioninterpretability analysisconvolutional neural network |
| spellingShingle | Jie Wu Yongjin He Chengyu Xu Xiaoping Jia Yule Huang Qianru Chen Chuyue Huang Armin Dadras Eslamlou Shiping Huang Interpretability Analysis of Convolutional Neural Networks for Crack Detection Buildings structural health monitoring crack detection interpretability analysis convolutional neural network |
| title | Interpretability Analysis of Convolutional Neural Networks for Crack Detection |
| title_full | Interpretability Analysis of Convolutional Neural Networks for Crack Detection |
| title_fullStr | Interpretability Analysis of Convolutional Neural Networks for Crack Detection |
| title_full_unstemmed | Interpretability Analysis of Convolutional Neural Networks for Crack Detection |
| title_short | Interpretability Analysis of Convolutional Neural Networks for Crack Detection |
| title_sort | interpretability analysis of convolutional neural networks for crack detection |
| topic | structural health monitoring crack detection interpretability analysis convolutional neural network |
| url | https://www.mdpi.com/2075-5309/13/12/3095 |
| work_keys_str_mv | AT jiewu interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection AT yongjinhe interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection AT chengyuxu interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection AT xiaopingjia interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection AT yulehuang interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection AT qianruchen interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection AT chuyuehuang interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection AT armindadraseslamlou interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection AT shipinghuang interpretabilityanalysisofconvolutionalneuralnetworksforcrackdetection |