混合型晶圓圖缺陷辨識容易受到重疊缺陷圖樣之間相互干擾的影響。分割與分類可作為辨識流程中的兩個步驟：分割先定位可能的缺陷圖樣區域，分類再驗證其對應類別。然而，若分別使用獨立的分割模型與分類模型，會增加可訓練模型參數的數量。雖然共享編碼器可以減少可訓練參數量，但分割與分類對特徵的需求不同，使用同一個共享編碼器可能造成特徵層級干擾。為了解決上述問題，本文提出一個具可切換批次正規化（switchable batch normalization, SwitchBN）的分割導引式分類架構，用於混合型晶圓圖缺陷辨識。此架構採用雙分支設計：分割分支先產生類別特定的提議遮罩（class-specific proposal masks），用以表示可能的缺陷圖樣區域；分類分支再驗證其對應類別是否存在。此設計使分類驗證聚焦於分割所定位的可能缺陷圖樣區域，並降低混合型缺陷中重疊缺陷圖樣所造成的干擾。
本文使用兩個公開影像資料集 MixedWM38 與 WM-811K 評估所提出的架構，並以完全匹配率（exact match ratio, EMR）作為主要效能指標。在 MixedWM38 資料集上，所提出架構達到 0.9700 ± 0.0041 的 EMR，而基於 Yan et al. (2023) 與 Li et al. (2025) 模型的結果分別為 0.9444 ± 0.0050 與 0.9520 ± 0.0080。在 WM-811K 測試集上，所提出架構達到 0.8440 ± 0.0744 的 EMR，而基於 Chiu and Chen (2021) 模型的結果為 0.7460 ± 0.0518。這些結果顯示，使用分割定位可能的缺陷圖樣區域、使用分類驗證其對應類別，並透過 SwitchBN 進行任務特定正規化，對混合型晶圓圖缺陷辨識是有效的。

Mixed-type wafer map defect recognition is being challenged by mutual interference among their overlapping defect patterns. Segmentation and classification can be employed as two steps for recognition: segmentation first designates possible defect pattern regions, and classification then identifies their associated classes. However, using independent segmentation and classification models increases the number of trainable model parameters. In contrast sharing an encoder reduces number of trainable parameters, however segmentation and classification have different feature requirements, and the one shared encoder approach may yield feature interference. To address these issues, this paper presents a segmentation-guided classification framework with switchable batch normalization (SwitchBN) for mixed-type wafer map defect recognition. The framework uses a dual-branch architecture: the segmentation branch first generates class-specific proposal masks that designate possible defect pattern regions, and the classification branch then identifies whether the corresponding classes are indeed present. This design focuses classification identification on possible defect pattern regions designated by segmentation and reduces interference from overlapping defect patterns in mixed-type defects.
Two public image datasets, MixedWM38 and WM-811K, are used to evaluate the proposed framework, and an exact match ratio (EMR) is used as the primary performance metric. On the former dataset, the proposed framework achieves an EMR of 0.9700 ± 0.0041, compared with those of 0.9444 ± 0.0050 and 0.9520 ± 0.0080 based on the models by Yan et al. (2023) and Li et al. (2025), respectively. The proposed framework on the latter dataset achieves an EMR of 0.8440 ± 0.0744, compared with that of 0.7460 ± 0.0518 based on the model by Chiu and Chen (2021). These results indicate that using segmentation guided classification to identify the corresponding classes, and applying SwitchBN-based task-specific normalization are effective for mixed-type wafer map defect recognition.

[1] T. Kim and K. Behdinan, “Advances in machine learning and deep learning applications towards wafer map defect recognition and classification: a review,” Journal of Intelligent Manufacturing, vol. 34, no. 8, pp. 3215–3247, 2023.
[2] L. Y. Hsieh and T.-J. Hsieh, “A throughput management system for semiconductor wafer fabrication facilities: Design, systems and implementation,” Processes, vol. 6, no. 2, p. 16, 2018.
[3] L. Mönch, J. W. Fowler, and S. J. Mason, Production planning and control for semiconductor wafer fabrication facilities: modeling, analysis, and systems. Springer Science & Business Media, 2012.
[4] M. Piao, C. H. Jin, J. Y. Lee, and J.-Y. Byun, “Decision tree ensemble-based wafer map failure pattern recognition based on radon transform-based features,” IEEE Transactions on Semiconductor Manufacturing, vol. 31, no. 2, pp. 250–257, 2018.
[5] S. Nag, D. Makwana, S. Mittal, C. K. Mohan et al., “Wafersegclassnet-a lightweight network for classification and segmentation of semiconductor wafer defects,” Computers in Industry, vol. 142, p. 103720, 2022.
[6] R. Wang and S. Wang, “Similarity searching for fault diagnosis of defect patterns in wafer bin maps,” Computers & Industrial Engineering, vol. 185, p. 109679, 2023.
[7] S. Yoon and S. Kang, “Semi-automatic wafer map pattern classification with convolutional neural networks,” Computers & Industrial Engineering, vol. 166, p. 107977, 2022.
[8] M. H. Hansen, V. N. Nair, and D. J. Friedman, “Monitoring wafer map data from integrated circuit fabrication processes for spatially clustered defects,” Technometrics, vol. 39, no. 3, pp. 241–253, 1997.
[9] G. Tello, O. Y. Al-Jarrah, P. D. Yoo, Y. Al-Hammadi, S. Muhaidat, and U. Lee,“Deep-structured machine learning model for the recognition of mixed-defect patterns in semiconductor fabrication processes,” IEEE Transactions on Semiconductor Manufacturing, vol. 31, no. 2, pp. 315–322, 2018.
[10] S. Chen, M. Liu, X. Hou, Z. Zhu, Z. Huang, and T. Wang, “Wafer map defect pattern detection method based on improved attention mechanism,” Expert Systems with Applications, vol. 230, p. 120544, 2023.
[11] C. Hansen and P. Thyregod, “Use of wafer maps in integrated circuit manufacturing,”Microelectronics Reliability, vol. 38, no. 6-8, pp. 1155–1164, 1998.
[12] K. W. Tobin Jr, S. S. Gleason, T. P. Karnowski, S. L. Cohen, and F. Lakhani,“Automatic classification of spatial signatures on semiconductor wafer maps,” in Metrology, Inspection, and Process Control for Microlithography XI, vol. 3050. SPIE, 1997, pp. 434–444.
[13] J. Y. Hwang and W. Kuo, “Model-based clustering for integrated circuit yield enhancement,” European Journal of Operational Research, vol. 178, no. 1, pp. 143–153, 2007.
[14] F.-L. Chen and S.-F. Liu, “A neural-network approach to recognize defect spatial pattern in semiconductor fabrication,” IEEE transactions on semiconductor manufacturing, vol. 13, no. 3, pp. 366–373, 2000.
[15] M.-J. Wu, J.-S. R. Jang, and J.-L. Chen, “Wafer map failure pattern recognition and similarity ranking for large-scale data sets,” IEEE Transactions on Semiconductor Manufacturing, vol. 28, no. 1, pp. 1–12, 2014.
[16] M. Saqlain, B. Jargalsaikhan, and J. Y. Lee, “A voting ensemble classifier for wafer map defect patterns identification in semiconductor manufacturing,” IEEE Transactions on Semiconductor Manufacturing, vol. 32, no. 2, pp. 171–182, 2019.
[17] T. Nakazawa and D. V. Kulkarni, “Wafer map defect pattern classification and image retrieval using convolutional neural network,” IEEE Transactions on Semiconductor Manufacturing, vol. 31, no. 2, pp. 309–314, 2018.
[18] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778.
[19] H.-C. Shao, C.-H. Chang, Y.-H. Lin, C.-W. Lin, S.-Y. Fang, and Y.-H. Liu, “Exploring “many in few＂and“few in many＂properties in long-tailed, highly-imbalanced ic defect classification,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2025.
[20] J. Choi and D. Suh, “A depthwise convolutional neural network model based on active contour for multi-defect wafer map pattern classification,” Engineering Applications of Artificial Intelligence, vol. 139, p. 109707, 2025.
[21] Y. Kong and D. Ni, “Qualitative and quantitative analysis of multi-pattern wafer bin maps,” IEEE Transactions on Semiconductor Manufacturing, vol. 33, no. 4, pp. 578–586, 2020.
[22] J. Wang, C. Xu, Z. Yang, J. Zhang, and X. Li, “Deformable convolutional networks for efficient mixed-type wafer defect pattern recognition,” IEEE Transactions on Semiconductor Manufacturing, vol. 33, no. 4, pp. 587–596, 2020.
[23] K. Kyeong and H. Kim, “Classification of mixed-type defect patterns in wafer bin maps using convolutional neural networks,” IEEE Transactions on Semiconductor Manufacturing, vol. 31, no. 3, pp. 395–402, 2018.
[24] Y. Wei and H. Wang, “Mixed-type wafer defect recognition with multi-scale information fusion transformer,” IEEE Transactions on Semiconductor Manufacturing, vol. 35, no. 2, pp. 341–352, 2022.
[25] A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly et al., “An image is worth 16x16 words: Transformers for image recognition at scale,” arXiv preprint arXiv:2010.11929, 2020.
[26] S. Chen, Z. Huang, T. Wang, X. Hou, and J. Ma, “Wafer map defect recognition based on multi-scale feature fusion and attention spatial pyramid pooling,” Journal of Intelligent Manufacturing, vol. 36, no. 1, pp. 271–284, 2025.
[27] H. Lee and H. Kim, “Semi-supervised multi-label learning for classification of wafer bin maps with mixed-type defect patterns,” IEEE Transactions on Semiconductor Manufacturing, vol. 33, no. 4, pp. 653–662, 2020.
[28] J. Wang, Z. Yang, J. Zhang, Q. Zhang, and W.-T. K. Chien, “Adabalgan: An improved generative adversarial network with imbalanced learning for wafer defective pattern recognition,” IEEE Transactions on Semiconductor Manufacturing, vol. 32, no. 3, pp. 310–319, 2019.
[29] T.-H. Tsai and Y.-C. Lee, “A light-weight neural network for wafer map classification based on data augmentation,” IEEE Transactions on Semiconductor Manufacturing, vol. 33, no. 4, pp. 663–672, 2020.
[30] K. S.-M. Li, X.-H. Jiang, L. L.-Y. Chen, S.-J. Wang, A. Y.-A. Huang, J. E. Chen, H.-C. Liang, and C.-L. Hsu, “Wafer defect pattern labeling and recognition using semi-supervised learning,” IEEE Transactions on Semiconductor Manufacturing, vol. 35, no. 2, pp. 291–299, 2022.
[31] J. Li, R. Tao, R. Chen, Y. Chen, C. Zhao, and X. Huang, “Sample-imbalanced wafer map defects classification based on auxiliary classifier denoising diffusion probability model,” Computers & Industrial Engineering, vol. 192, p. 110209, 2024.
[32] W. Shin, H. Kahng, and S. B. Kim, “Mixup-based classification of mixed-type defect patterns in wafer bin maps,” Computers & Industrial Engineering, vol. 167, p. 107996, 2022.
[33] J. Shim and S. Kang, “Learning from single-defect wafer maps to classify mixed-defect wafer maps,” Expert Systems with Applications, vol. 233, p. 120923, 2023.
[34] N. D. Thuan, “Inspection of mixed-type wafer defects from single-defect data via adaptive synthesis and self-attention cnn,” Measurement, vol. 256, p. 118306, 2025.
[35] M.-C. Chiu and T.-M. Chen, “Applying data augmentation and mask r-cnn-based instance segmentation method for mixed-type wafer maps defect patterns classification,” IEEE Transactions on Semiconductor Manufacturing, vol. 34, no. 4, pp. 455–463, 2021.
[36] D. Tabernik, S. Šela, J. Skvarč, and D. Skočaj, “Segmentation-based deep-learning approach for surface-defect detection,” Journal of Intelligent Manufacturing, vol. 31, no. 3, pp. 759–776, 2020.
[37] J. Yan, Y. Sheng, and M. Piao, “Semantic segmentation-based wafer map mixed-type defect pattern recognition,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 42, no. 11, pp. 4065–4074, 2023.
[38] J. Li, K. Long, R. Chen, Y. Li, and X. Huang, “Mixed-defect wafer map separation and detection based on single-defect wafer map,” Computers & Industrial Engineering, p. 111395, 2025.
[39] S. Yin, Y. Zhang, and R. Wang, “Contrastive learning with global and local representation for mixed-type wafer defect recognition,” Sensors, vol. 25, no. 4, p. 1272, 2025.
[40] Y. Zhang and Q. Yang, “A survey on multi-task learning,” IEEE transactions on knowledge and data engineering, vol. 34, no. 12, pp. 5586–5609, 2021.
[41] S. Vandenhende, S. Georgoulis, W. Van Gansbeke, M. Proesmans, D. Dai, and L. Van Gool, “Multi-task learning for dense prediction tasks: A survey,” IEEE transactions on pattern analysis and machine intelligence, vol. 44, no. 7, pp. 3614–3633, 2021.
[42] T. Yu, S. Kumar, A. Gupta, S. Levine, K. Hausman, and C. Finn, “Gradient surgery for multi-task learning,” Advances in neural information processing systems, vol. 33, pp. 5824–5836, 2020.
[43] A. Navon, A. Shamsian, I. Achituve, H. Maron, K. Kawaguchi, G. Chechik, and E. Fetaya, “Multi-task learning as a bargaining game,” arXiv preprint arXiv:2202.01017, 2022.
[44] Y. Li, N. Wang, J. Shi, J. Liu, and X. Hou, “Revisiting batch normalization for practical domain adaptation,” arXiv preprint arXiv:1603.04779, 2016.
[45] W.-G. Chang, T. You, S. Seo, S. Kwak, and B. Han, “Domain-specific batch normalization for unsupervised domain adaptation,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition, 2019, pp. 7354–7362.
[46] C. Xie, M. Tan, B. Gong, J. Wang, A. L. Yuille, and Q. V. Le, “Adversarial examples improve image recognition,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 819–828.
[47] C. H. Sudre, W. Li, T. Vercauteren, S. Ourselin, and M. Jorge Cardoso, “Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations,” in International Workshop on Deep Learning in Medical Image Analysis. Springer, 2017, pp. 240–248.
[48] Z. Zhou, M. M. Rahman Siddiquee, N. Tajbakhsh, and J. Liang, “Unet++: A nested u-net architecture for medical image segmentation,” in International workshop on deep learning in medical image analysis. Springer, 2018, pp. 3–11.
[49] O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in International Conference on Medical image computing and computer-assisted intervention. Springer, 2015, pp. 234–241.
[50] T.-Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 2117–2125.
[51] L.-C. Chen, Y. Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder-decoder with atrous separable convolution for semantic image segmentation,” in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 801–818.