Deformable Convolutional Networks for Efficient Mixed-Type Wafer Defect Pattern Recognition
The DOI matches its Crossref record; 35 of the 35 references it cites were found in a registry or an index.
Abstract
Defect pattern recognition (DPR) of wafer maps is critical for determining the root cause of production defects, which can provide insights for the yield improvement in wafer foundries. During wafer fabrication, several types of defects can be coupled together in a piece of wafer, it is called mixed-type defects DPR. To detect mixed-type defects is much more complicated because the combination of defects may vary a lot, from the type of defects, position, angle, number of defects, etc. Deep learning methods have been a good choice for complex pattern recognition problems. In this article, we propose a deformable convolutional network (DC-Net) for mixed-type DPR (MDPR) in which several types of defects are coupled together in a piece of wafer. A deformable convolutional unit is designed to selectively sample from mixed defects, then extract high-quality features from wafer maps. A multi-label output layer is improved with a one-hot encoding mechanism, which decomposes extract mixed features into each basic single defect. The experiment results indicate that the proposed DC-Net model outperforms conventional models and other deep learning models. Further results of the interpretable analysis reveal that the proposed DC-Net can accurately pinpoint the defects areas of wafer maps with noise points, which is beneficial for mixed-type DPR problems.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
Deformable Convolutional Networks
Dai, J., Qi, H., Xiong, Y., Li, Y., Zhang, G., Hu, H., & Wei, Y. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.89 · Cited by 7,513 · cited 1×
“Among the machine learning technologies, the deformable convolutional (DC) network [27] selectively samples from input data based on the position information for further displacement adjustment, which has been extensively applied in multi-angle object detection and hyperspectral image classification…”
BackgroundA new convolutional neural network-based data-driven fault diagnosis method
Wen, L., Li, X., Gao, L., & Zhang, Y. (2018) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2017.2774777 · Cited by 2,222 · cited 1×
“With the development of artificial intelligence, deep learning methods extract features from raw data automatically through multi-layer neural networks [21]–[23], which shows incredible performance in computer version tasks [24].”
MethodSelf-citationEdvr: video restoration with enhanced deformable convolutional networks
Wang, X., Chan, K. C. K., Yu, K., Dong, C., & Loy, C. C. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw.2019.00247 · Cited by 1,272 · cited 1×
“Wang et al. [31] devised a DC neural network module to align different frames at the feature level to handle large motions in video restoration tasks.”
BackgroundLee, K. B., Cheon, S., & Kim, C. O. (2017) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2017.2676245 · Cited by 483 · cited 1×
“After the wafer manufacturing, wafer test is conducted to evaluate the electrical function of each die by probe test [2], which consists of many test items that measure the performance of the operations of required die electronic functions.”
MethodWafer Map Defect Pattern Classification and Image Retrieval Using Convolutional Neural Network
Nakazawa, T., & Kulkarni, D. V. (2018) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2018.2795466 · Cited by 408 · cited 3×
“The performances of the proposed DC-Net were first compared with Takeshi_CNN [25], Kiryong_CNN [26].”
BackgroundMethodWafer Map Failure Pattern Recognition and Similarity Ranking for Large-Scale Data Sets
Wu, M.-J., Jang, J.-S. R., & Chen, J.-L. (2015) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2014.2364237 · Cited by 400 · cited 2×
“Aiming to achieve high accuracy, shallow models should be equipped with good feature extractors (such as correlogram and Radon transform [18]), which can obtain differentiable features from raw data.”
BackgroundA brief introduction to OpenCV
Čuljak, I., Abram, D., Pribanić, T., Džapo, H., & Cifrek, M. (2012) · International Convention on Information and Communication Technology, Electronics and Microelectronics · Cited by 376
AI-based modeling and data-driven evaluation for smart manufacturing processes
Ghahramani, M., Qiao, Y., Zhou, M. C., O'Hagan, A., & Sweeney, J. (2020) · IEEE/CAA Journal of Automatica Sinica · doi:10.1109/jas.2020.1003114 · Cited by 334 · cited 1×
“With the development of artificial intelligence, deep learning methods extract features from raw data automatically through multi-layer neural networks [21]–[23], which shows incredible performance in computer version tasks [24].”
MethodDeformable Convolutional Neural Networks for Hyperspectral Image Classification
Zhu, J., Fang, L., & Ghamisi, P. (2018) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2018.2830403 · Cited by 244 · cited 1×
“…the machine learning technologies, the deformable convolutional (DC) network [27] selectively samples from input data based on the position information for further displacement adjustment, which has been extensively applied in multi-angle object detection and hyperspectral image classification [28].”
BackgroundClassification of Mixed-Type Defect Patterns in Wafer Bin Maps Using Convolutional Neural Networks
Kyeong, K., & Kim, H. (2018) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2018.2841416 · Cited by 228 · cited 9×
“In Kyeong and Kim’s work [26], 16 defects mixed with”
BackgroundMethodInfluentialMönch, L., Uzsoy, R., & Fowler, J. W. (2018) · International Journal of Production Research · doi:10.1080/00207543.2017.1401233 · Cited by 168 · cited 1×
“T HE INTEGRATED CIRCUIT (IC) industry consists of four basic parts: IC design, wafer manufacturing, wafer test, assembly, and final test [1], as is shown in Fig.”
MethodDeep imitation learning for autonomous vehicles based on convolutional neural networks
Kebria, P. M., Khosravi, A., Salaken, S. M., & Nahavandi, S. (2020) · IEEE/CAA Journal of Automatica Sinica · doi:10.1109/jas.2019.1911825 · Cited by 161
Wang, J., Yang, Z., Zhang, J., Zhang, Q., & Chien, W.-T. K. (2019) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2019.2925361 · Cited by 154 · cited 1×
“The DPR of wafer maps is a computer vision task, which has received sufficient consideration [10], [11].”
BackgroundSelf-citationDetection of Spatial Defect Patterns Generated in Semiconductor Fabrication Processes
Yuan, T., Kuo, W., & Bae, S. J. (2011) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2011.2154870 · Cited by 143 · cited 1×
“Yuan et al. [20] proposed a multistep defect analysis approach to recognize six mixed-type defective patterns, which provided a similarity-based clustering technique to group the local defects into clusters according to their spatial locations.”
MethodChien, C., Hsu, C., & Chen, P.-N. (2012) · Flexible Services and Manufacturing Journal · doi:10.1007/s10696-012-9161-4 · Cited by 142 · cited 1×
“Defect pattern recognition (DPR) classifies wafer maps into several categories by analyzing the distribution of dies failed in wafer test, which provides insights for reasoning the root cause of defects in fabrication processes [5], [6].”
BackgroundA collaborative architecture of the industrial internet platform for manufacturing systems
Wang, J., Xu, C., Zhang, J., Bao, J., & Zhong, R. Y. (2019) · Robotics and Computer-Integrated Manufacturing · doi:10.1016/j.rcim.2019.101854 · Cited by 133 · cited 1×
“Recently, with the development of intelligent manufacturing technologies [16], [17], data-driven methods have been widely applied in the mixed-type DPR.”
MethodSelf-citationTemporal Deformable Convolutional Encoder-Decoder Networks for Video Captioning
Chen, J., Pan, Y., Li, Y., Yao, T., Chao, H., & Mei, T. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.33018167 · Cited by 122
Cao, D., Chen, Z., & Gao, L. (2020) · Human-centric Computing and Information Sciences · doi:10.1186/s13673-020-00219-9 · Cited by 114 · cited 1×
“Cao et al. [30] proposed a multi-scaled DC object detection network to detect small, dense objects.”
BackgroundDefect spatial pattern recognition using a hybrid SOM–SVM approach in semiconductor manufacturing
Li, T.-S., & Huang, C.-L. (2007) · Expert Systems with Applications · doi:10.1016/j.eswa.2007.09.023 · Cited by 99 · cited 1×
“The DPR of wafer maps is a computer vision task, which has received sufficient consideration [10], [11].”
BackgroundWang, J., Zhang, J., & Wang, X. (2018) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2017.2788501 · Cited by 95 · cited 2×
“Directed by the Moore’s law, the layer of the integrated circuits and the number of the dies on a wafer are increased greatly with the development of integrated circuits technology [9].”
BackgroundSelf-citationDetection and clustering of mixed-type defect patterns in wafer bin maps
Kim, J.-H., Lee, Y., & Kim, H. (2018) · IISE Transactions · doi:10.1080/24725854.2017.1386337 · Cited by 93 · cited 2×
“…methods can be divided into two types: model-based methods and data-driven methods. the model-based methods use a predefined defective function (such as probability distribution function or spatial dependence function) for each wafer map and select the best matching defect pattern [12], [13].”
MethodAdly, F., Alhussein, O., Yoo, P. D., Al-Hammadi, Y., Taha, K., Muhaidat, S., Jeong, Y., Lee, U.-H., & Ismail, M. (2015) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2015.2481719 · Cited by 92 · cited 3×
“Shallow models refer to the machine learning models with shallow structures, such as the RBFN, decision forest, and support vector machine [5].”
BackgroundMethodTaam, W., & Hamada, M. S. (1993) · Technometrics · doi:10.1080/00401706.1993.10485037 · Cited by 75 · cited 1×
“…methods can be divided into two types: model-based methods and data-driven methods. the model-based methods use a predefined defective function (such as probability distribution function or spatial dependence function) for each wafer map and select the best matching defect pattern [12], [13].”
MethodAccurate and robust eye center localization via fully convolutional networks
Xia, Y., Yu, H. L., & Wang, F. (2019) · IEEE/CAA Journal of Automatica Sinica · doi:10.1109/jas.2019.1911684 · Cited by 68 · cited 1×
“With the development of artificial intelligence, deep learning methods extract features from raw data automatically through multi-layer neural networks [21]–[23], which shows incredible performance in computer version tasks [24].”
MethodFog-IBDIS: Industrial Big Data Integration and Sharing with Fog Computing for Manufacturing Systems
Wang, J., Zheng, P., Lv, Y., Bao, J., & Zhang, J. (2019) · Engineering · doi:10.1016/j.eng.2018.12.013 · Cited by 66 · cited 1×
“Recently, with the development of intelligent manufacturing technologies [16], [17], data-driven methods have been widely applied in the mixed-type DPR.”
MethodSelf-citationUsing Wafer Map Features to Better Predict Die-Level Failures in Final Test
Kang, S., Cho, S., An, D., & Rim, J. (2015) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2015.2443864 · Cited by 65 · cited 1×
“The manual defective pattern recognition can hardly satisfy the system needs [7].”
BackgroundLiu, F., Liu, D., Tian, J., Xie, X., Yang, X., & Wang, K. (2020) · Medical Image Analysis · doi:10.1016/j.media.2020.101793 · Cited by 56 · cited 1×
“A one-shot DC module was designed with a predictor-net to predicts interpolation offsets for guiding the deformation of the receptive field to enhance the robustness of respiratory motion estimation in ultrasound sequences [29].”
MethodSeparation of composite defect patterns on wafer bin map using support vector clustering
Wang, C.-H. (2008) · Expert Systems with Applications · doi:10.1016/j.eswa.2008.01.057 · Cited by 49 · cited 1×
“Wang [19] integrated classifier chains and support vector machines (CC-SVM) together to identify seven mixed-type defective patterns.”
BackgroundChao, L.-C., & Tong, L. (2009) · Expert Systems with Applications · doi:10.1016/j.eswa.2009.01.003 · Cited by 48 · cited 1×
“Data-driven methods classify wafer maps through machine learning from massive data [14], [15], such as the RBFN, decision forest, and support vector machine [5].”
MethodPractical "building-in reliability" approaches for semiconductor manufacturing
Chen, W., & Huang, C. H.-J. (2002) · IEEE Transactions on Reliability · doi:10.1109/tr.2002.804494 · Cited by 43 · cited 1×
“Then, these qualified dies are cut from the wafer, assumed into individual chips, and label as qualified products after the final test [3].”
BackgroundMulti-step ART1 algorithm for recognition of defect patterns on semiconductor wafers
Choi, G., Kim, S.-H., Ha, C., & Bae, S. J. (2011) · International Journal of Production Research · doi:10.1080/00207543.2011.574502 · Cited by 40 · cited 1×
“Data-driven methods classify wafer maps through machine learning from massive data [14], [15], such as the RBFN, decision forest, and support vector machine [5].”
MethodFuzzy neural network-based rescheduling decision mechanism for semiconductor manufacturing
Zhang, J., Qin, W., Wu, L. H., & Zhai, W. (2014) · Computers in Industry · doi:10.1016/j.compind.2014.06.002 · Cited by 37 · cited 1×
“Therefore, there is a great demand for automated DPR in the operation of SWFSs. Duringthe wafer fabrication, the integrated circuits were built layer-by-layer in each die on a piece of silicon wafer [8].”
BackgroundSelf-citationTaam, W., & Hamada, M. S. (1993) · Technometrics · doi:10.2307/1269660 · Cited by 35
Gesture Recognition Using Spatiotemporal Deformable Convolutional Representation
Shi, L., Zhang, Y., Hu, J. F., Cheng, J., & Lu, H. (2019) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2019.8803152 · Cited by 26 · cited 1×
“A novel spatiotemporal DC neural network [32] was designed for end-to-end learning to extract complex features for dynamic gesture recognition.”
BackgroundElectromigration Lifetime Optimization by Uniform Designs and a New Lifetime Index
Yang, S. F., & Chien, W.-T. K. (2015) · IEEE Transactions on Reliability · doi:10.1109/tr.2015.2453117 · Cited by 12 · cited 1×
“The testing results of all dies on a wafer are saved on a wafer map [4], which has a two-dimensional adjacent matrix.”
Method
Cited by
Big data analytics for intelligent manufacturing systems: A review
Wang, J., Xu, C., Zhang, J., & Zhong, R. (2022) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2021.03.005 · Cited by 603
Self-citationA Review on Recent Advances in Vision-based Defect Recognition towards Industrial Intelligence
Gao, Y., Li, X., Wang, X. V., Wang, L., & Gao, L. (2022) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2021.05.008 · Cited by 253
Self-citationKim, T., & Behdinan, K. (2022) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-022-01994-1 · Cited by 175
“The related works for mixed-type WM defect detection framed the problem as multi-label classification (Devika & George, 2019; Hyun and Kim (2020); Wang et al., 2020) or multi-class classification (Byun & Baek, 2020; Kim et al., 2021; Kong & Ni, 2019, 2020b; Kyeong & Kim, 2018; Zhuang et al., 2020).”
BackgroundMethodInfluentialNag, S., Makwana, D., Sai Chandra Teja, R., Mittal, S., & Mohan, C. K. (2022) · Computers in Industry · doi:10.1016/j.compind.2022.103720 · Cited by 98
“For defect classification, we compare our model with ResNet50 [11], DenseNet121 [12], LeNet [13], AlexNet [14], EfficientNetB0 [15], MobileNetV2 [16], ResNet18 [11], and DCNet [8], a recently proposed CNN for wafer defect classification.”
BackgroundMethodResultInfluentialA Systematic Review of Deep Learning for Silicon Wafer Defect Recognition
Batool, U., Shapiai, M. I., Tahir, M., Ismail, Z. H., Zakaria, N. J., & Elfakharany, A. (2021) · IEEE Access · doi:10.1109/access.2021.3106171 · Cited by 106
“Wang et al. [59] proposed a deformable convolutional network for mixed-type defect patterns by selectively sampling and extracting high-quality features from mixed wafer defects.”
Background
Metadata
- Authors
- Wang, J., Xu, C., Yang, Z., Zhang, J., & Li, X.
- Affiliations
- Donghua University (CN)
- Shanghai Jiao Tong University (CN)
- Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional (MX)
- Venue
- IEEE Transactions on Semiconductor Manufacturing
- Published
- 2020-11
- Volume
- 33
- Issue
- 4
- Pages
- 587-596
- ISSN
- 0894-6507 (ISSN-L)
- 1558-2345
- OpenAlex
- W3082906739
- Cited by
- 229
- References
- 34
- Citations by year
- 2026: 53 · 2025: 52 · 2024: 42 · 2023: 38 · 2022: 20 · 2021: 22 · 2020: 2
- FWCI
- 13.07
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Transactions on Semiconductor Manufacturing (publishedVersion)
- Funders
- National Natural Science Foundation of China (51905091)
- Fundamental Research funds for the Central Universities (2232019D3-34)
- Shanghai Sailing Program (19YF1401500)
- Topics
- Physical Sciences › Engineering › Industrial and Manufacturing Engineering › Industrial Vision Systems and Defect Detection
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Engineering › Control and Systems Engineering › Fault Detection and Control Systems
- Keywords
- deformable convolutional networks, wafer maps, multi-label classification, one-hot encoding, defect root cause analysis, explainable deep learning
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.