An Efficient Convolutional Neural Network Model Based on Object-Level Attention Mechanism for Casting Defect Detection on Radiography Images
The DOI matches its Crossref record; 46 of the 46 references it cites were found in a registry or an index.
Abstract
Automatic detection of casting defects on radiography images is an important technology to automatize digital radiography defect inspection. Traditionally, in an industrial application, conventional methods are inefficient when the detection targets are small, local, and subtle in the complex scenario. Meanwhile, the outperformance of deep learning models, such as the convolutional neural network (CNN), is limited by a huge volume of data with precise annotations. To overcome these challenges, an efficient CNN model, only trained with image-level labels, is first proposed for detection of tiny casting defects in a complicated industrial scene. Then, in this article, we present a novel training strategy which can form a new object-level attention mechanism for the model during the training phase, and bilinear pooling is utilized to improve the model capability of detecting local contrast casting defects. Moreover, to enhance the interpretability, we extend class activation maps (CAM) to bilinear CAM (Bi-CAM) which is adapted to bilinear architectures as a visualization technique to reason about the model output. Experimental results show that the proposed model achieves superior performance in terms of each quantitative metric and is suitable for most actual applications. The real-time defect detection of castings is efficiently implemented in the complex scenario.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“We implement the TCM based on VGG16 [44] which is effective and light.”
MethodImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 1×
“The same training details are: 1) initial parameters pretrained on the ImageNet [46]; 2) training with 40 epochs; and 3) optimizer which uses stochastic gradient descent (SGD) with 0.9 momentum parameter and 0.0005 weight decay.”
MethodInfluentialSqueeze-and-Excitation Networks
Hu, J., Shen, L., & Sun, G. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00745 · Cited by 35,667
CBAM: Convolutional Block Attention Module
Woo, S., Park, J., Lee, J.-Y., & Kweon, I. S. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-01234-2_1 · Cited by 25,853
Xception: Deep Learning with Depthwise Separable Convolutions
Chollet, F. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.195 · Cited by 20,380 · cited 2×
“Duetotheefficiencyof depthwiseseparableconvolution[26],wereplaceentirestandard convolutions to the depthwise separable convolutions in residual blocks.”
BackgroundMethodSqueeze-and-Excitation Networks
Hu, J., Shen, L., Albanie, S., Sun, G., & Wu, E. (2020) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2019.2913372 · Cited by 12,157
Learning Deep Features for Discriminative Localization
Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., & Torralba, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.319 · Cited by 11,428 · cited 3×
“3) Inspiredbyclassactivationmaps(CAM)[27],wepropose bilinearCAM(Bi-CAM),whichissuitabletobilinear architectures,tobeusedasavisualizationtechniqueto increasetheinterpretabilityofthemodel.”
BackgroundMethodInfluentialA model of saliency-based visual attention for rapid scene analysis
Itti, L., Koch, C., & Niebur, E. (1998) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/34.730558 · Cited by 11,271 · cited 1×
“In recent years, inspired by human perception process [37], attention mechanism has extensive applications [18]–[21].”
BackgroundShow, Attend and Tell: Neural Image Caption Generation with Visual Attention
Xu, K., Ba, J., Kiros, R., Cho, K., Courville, A., Salakhutdinov, R., Zemel, R., & Bengio, Y. (2015) · arXiv · doi:10.48550/arxiv.1502.03044 · Cited by 8,248 · cited 1×
“Xu et al. [38] proposeanattention-basedmodelforimagecaptiontasks.”
(2011) · Cited by 6,993
Residual Attention Network for Image Classification
Wang, F., Jiang, M., Qian, C., Yang, S., Li, C., Zhang, H., Wang, X., & Tang, X. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.683 · Cited by 4,073 · cited 1×
“Furthermore, some special modules abstract the attention concept from spatial level to channel level [42].”
BackgroundRecent developments in advanced aircraft aluminium alloys
Dursun, T., & Soutis, C. (2014) · Materials & Design (1980-2015) · doi:10.1016/j.matdes.2013.12.002 · Cited by 2,807 · cited 1×
“A LUMINIUM castings have been extensively applied to various parts of automobile [1], aerospace [2], aircraft [3], and electric devices [4] whose qualities affect the fatigue behavior of the overall product [5].”
BackgroundRecent development in aluminium alloys for the automotive industry
Miller, W. S., Zhuang, L., Bottema, J., Wittebrood, A. J., De Smet, P., Haszler, A., & Vieregge, A. (2000) · Materials Science and Engineering: A · doi:10.1016/s0921-5093(99)00653-x · Cited by 2,498 · cited 1×
“A LUMINIUM castings have been extensively applied to various parts of automobile [1], aerospace [2], aircraft [3], and electric devices [4] whose qualities affect the fatigue behavior of the overall product [5].”
BackgroundHarmonious Attention Network for Person Re-identification
Li, W., Zhu, X., & Gong, S. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00243 · Cited by 1,449
Recent development in aluminium alloys for aerospace applications
Heinz, A., Haszler, A., Keidel, C., Moldenhauer, S., Benedictus, R., & Miller, W. S. (2000) · Materials Science and Engineering A · doi:10.1016/s0921-5093(99)00674-7 · Cited by 1,357
Guo, L., Lei, Y., Xing, S., Yan, T., & Li, N. (2019) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2018.2877090 · Cited by 1,234 · cited 1×
“Different from transfer learning where the existing knowledge is learned to solve a new but related task [43], the core perspective of our training strategy is to train the model”
BackgroundMulti-modal Factorized Bilinear Pooling with Co-attention Learning for Visual Question Answering
Yu, Z., Yu, J., Fan, J., & Tao, D. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.202 · Cited by 758
Riera-Guasp, M., Antonino-Daviu, J. A., & Capolino, G.-A. (2015) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2014.2375853 · Cited by 573 · cited 1×
“A LUMINIUM castings have been extensively applied to various parts of automobile [1], aerospace [2], aircraft [3], and electric devices [4] whose qualities affect the fatigue behavior of the overall product [5].”
BackgroundSkeleton-Based Human Action Recognition With Global Context-Aware Attention LSTM Networks
Liu, J., Wang, G., Duan, L.-Y., Abdiyeva, K., & Kot, A. C. (2018) · IEEE Transactions on Image Processing · doi:10.1109/tip.2017.2785279 · Cited by 537
Skeleton-Based Human Action Recognition with Global Context-Aware Attention LSTM Networks
Liu, J., Wang, G., Duan, L.-Y., Abdiyeva, K., & Kot, A. C. (2018) · Lancaster EPrints (Lancaster University) · Cited by 531
Developments in investment casting process—A review
Pattnaik, S., Karunakar, D. B., & Jha, P. K. (2012) · Journal of Materials Processing Technology · doi:10.1016/j.jmatprotec.2012.06.003 · Cited by 501 · cited 1×
“Digital Object Identifier 10.1109/TIE.2019.2962437 of the casting process [6], defects are inevitable during the process, such as gas holes, sand holes, and flaws.”
BackgroundInfluentialTian, J., Morillo, C., Azarian, M. H., & Pecht, M. (2016) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2015.2509913 · Cited by 488 · cited 1×
“Machine learning-based methods have been proposed more often, which utilize feature algorithms to [12],[13].”
MethodAutomatic fruit classification using deep learning for industrial applications
Hossain, M. S., Al-Hammadi, M., & Muhammad, G. (2019) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2018.2875149 · Cited by 347 · cited 1×
“Recently, deep learning model, particularly the convolutional neural network (CNN), has received substantial interest in industrial applications [14]–[17].”
BackgroundBilinear Convolutional Neural Networks for Fine-Grained Visual Recognition
Lin, T.-Y., RoyChowdhury, A., & Maji, S. (2018) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2017.2723400 · Cited by 328 · cited 2×
“With the success of bilinear pooling in fine-grained image classification tasks [22], the bilinear pooling has been utilized as a common method to obtain deeper and superior feature representation [23]–[25].”
MethodThree-Stream Attention-Aware Network for RGB-D Salient Object Detection
Chen, H., & Li, Y. (2019) · IEEE Transactions on Image Processing · doi:10.1109/tip.2019.2891104 · Cited by 324 · cited 2×
“Attention mechanism, which is able to select more interested attributes of input data, has been proposed and successfully used in many different tasks [18]–[21].”
BackgroundMethodCheon, S., Lee, H., Kim, C. O., & Lee, S. H. (2019) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2019.2902657 · Cited by 304 · cited 1×
“[34] design a CNN to classify defects of wafer surface and utilize feature vectors extracted by the network to construct a k-NN model for classification”
MethodMachine vision intelligence for product defect inspection based on deep learning and Hough transform
Wang, J., Fu, P., & Gao, R. X. (2019) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2019.03.002 · Cited by 288 · cited 1×
“In order to obtain the internalinformationandguaranteethecompletenessofcastings, radiography is often used for nondestructive testing [7], which has been widely used in quality controlling [8] and security inspection system [9].”
MethodDeep Attention-Based Spatially Recursive Networks for Fine-Grained Visual Recognition
Wu, L., Wang, Y., Li, X., & Gao, J. (2019) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2018.2813971 · Cited by 234 · cited 1×
“With the success of bilinear pooling in fine-grained image classification tasks [22], the bilinear pooling has been utilized as a common method to obtain deeper and superior feature representation [23]–[25].”
MethodSelf-citationFine-grained attention mechanism for neural machine translation
Choi, H., Cho, K., & Bengio, Y. (2018) · Neurocomputing · doi:10.1016/j.neucom.2018.01.007 · Cited by 205
Modern Computer Vision Techniques for X-Ray Testing in Baggage Inspection
Mery, D., Svec, E., Arias, M., Riffo, V., Saavedra, J. M., & Banerjee, S. (2017) · IEEE Transactions on Systems, Man, and Cybernetics: Systems · doi:10.1109/tsmc.2016.2628381 · Cited by 188 · cited 1×
“In order to obtain the internalinformationandguaranteethecompletenessofcastings, radiography is often used for nondestructive testing [7], which has been widely used in quality controlling [8] and security inspection system [9].”
MethodAutomated detection of welding defects in pipelines from radiographic images DWDI
Boaretto, N., & Centeno, T. M. (2017) · NDT & E International · doi:10.1016/j.ndteint.2016.11.003 · Cited by 181 · cited 2×
“construct sample set and execute the detection inference from the statistical aspect [12], [13].”
BackgroundYang, H., Chen, Y., Song, K., & Yin, Z. (2019) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2018.2886031 · Cited by 176 · cited 1×
“Recently, deep learning model, particularly the convolutional neural network (CNN), has received substantial interest in industrial applications [14]–[17].”
BackgroundAttention Mechanisms in Visual Search—An fMRI Study
Leonards, U., Sunaert, S., Van Hecke, P., & Orban, G. A. (2000) · Journal of Cognitive Neuroscience · doi:10.1162/089892900564073 · Cited by 168 · cited 1×
“of vision [41].”
Content-based image retrieval with compact deep convolutional features
Alzu’bi, A., Amira, A., & Ramzan, N. (2017) · Neurocomputing · doi:10.1016/j.neucom.2017.03.072 · Cited by 146
Hangai, Y., Takahashi, K., Yamaguchi, R., Utsunomiya, T., Kitahara, S., Kuwazuru, O., & YOSHIKAWA, N. (2012) · Materials Science and Engineering A · doi:10.1016/j.msea.2012.07.047 · Cited by 135
Shanmugamani, R., Sadique, M., & Ramamoorthy, B. (2014) · Measurement · doi:10.1016/j.measurement.2014.10.009 · Cited by 133
Multiview Generative Adversarial Network and Its Application in Pearl Classification
Xuan, Q., Chen, Z., Liu, Y., Huang, H., Bao, G., & Zhang, D. (2018) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2018.2885684 · Cited by 119
Automatic fabric defect detection with a wide-and-compact network
Li, Y., Zhang, D., & Lee, D.-J. (2018) · Neurocomputing · doi:10.1016/j.neucom.2018.10.070 · Cited by 117 · cited 1×
“Awide-and-compactCNNarchitectureisconstructed byLi et al. [33] that can learn multiscale and multiposition features from fabric, and achieve the automatic detection of fabric defects.”
BackgroundMultisensor Fusion System for Monitoring High-Power Disk Laser Welding Using Support Vector Machine
You, D., Gao, X., & Katayama, S. (2014) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2014.2309482 · Cited by 116 · cited 1×
“Over the past decades, conventional computer vision technology has extensive applications in many automatic inspection systems [10]–[13].”
BackgroundAutomatic thresholding for defect detection by background histogram mode extents
Aminzadeh, M., & Kurfess, T. R. (2015) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2015.09.004 · Cited by 67 · cited 1×
“In computing-based methods, numerical information of images, such as color [28]–[30] and frequency [31], [32], is utilized to segment the defect regions from original images.”
MethodDefect detection in magnetic tile images based on stationary wavelet transform
Yang, C., Liu, P., Yin, G., Jiang, H., & Li, X. (2016) · NDT & E International · doi:10.1016/j.ndteint.2016.04.006 · Cited by 63 · cited 1×
“In computing-based methods, numerical information of images, such as color [28]–[30] and frequency [31], [32], is utilized to segment the defect regions from original images.”
MethodA robust vision inspection system for detecting surface defects of film capacitors
Yang, Y., Zha, Z.-J., Gao, M., & He, Z. (2015) · Signal Processing · doi:10.1016/j.sigpro.2015.10.028 · Cited by 44 · cited 1×
“In computing-based methods, numerical information of images, such as color [28]–[30] and frequency [31], [32], is utilized to segment the defect regions from original images.”
MethodAnwar, A., Lin, W., Deng, X., Qiu, J., & Gao, H. (2018) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2018.2881948 · Cited by 32 · cited 1×
“In computing-based methods, numerical information of images, such as color [28]–[30] and frequency [31], [32], is utilized to segment the defect regions from original images.”
MethodPixel-Wise Defect Detection by CNNs without Manually Labeled Training Data
Haselmann, M., & Gruber, D. P. (2019) · Applied Artificial Intelligence · doi:10.1080/08839514.2019.1583862 · Cited by 31 · cited 1×
“Haselmann et al. [35] construct an autoencoder by CNN and inject random artificial defects into samples to train the network to detect locally confined defects on fabric surface without manual labels.”
MethodA Novel Generative Model With Bounded-GAN for Reliability Classification of Gear Safety
Li, J., He, H., Li, L., et al. (2019) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2018.2889629 · Cited by 24 · cited 1×
“Li et al. [36] develop a new generative model and mean-covariance labeling scheme to expand the data space.”
BackgroundDynamical performance analysis of communication-embedded neural networks: A survey
Chen, W., Ding, D., Mao, J., Liu, H., & Hou, N. (2019) · Neurocomputing · doi:10.1016/j.neucom.2018.08.088 · Cited by 14
Cited by
A Semantic-Enhanced Method Based On Deep SVDD for Pixel-Wise Anomaly Detection
Hu, C., Chen, K., & Shao, H. (2021) · 2021 IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme51207.2021.9428370 · Cited by 24
“In many practical systems [10, 11, 12], localizing anomalous regions is valuable.”
BackgroundSelf-citation
Metadata
- Authors
- Hu, C., & Wang, Y.
- Venue
- IEEE Transactions on Industrial Electronics
- Published
- 2020-12
- Volume
- 67
- Issue
- 12
- Pages
- 10922-10930
- ISSN
- 0278-0046 (ISSN-L)
- 1557-9948
- OpenAlex
- W2998008435
- Cited by
- 159
- References
- 46
- FWCI
- 14.38
- Type
- journal article
- Language
- en
- Open access
- closed
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.