제조 공정 결함 탐지를 위한 MixMatch 기반 준지도학습 성능 분석
Performance Analysis of MixMatch-Based Semi-Supervised Learning for Defect Detection in Manufacturing Processes
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Recently, there has been an increasing attempt to replace defect detection inspections in the manufacturing industry using deep learning techniques. However, obtaining substantial high-quality labeled data to enhance the performance of deep learning models entails economic and temporal constraints. As a solution for this problem, semi-supervised learning, using a limited amount of labeled data, has been gaining traction. This study assesses the effectiveness of semi-supervised learning in the defect detection process of manufacturing using the MixMatch algorithm. The MixMatch algorithm incorporates three dominant paradigms in the semi-supervised field: Consistency regularization, Entropy minimization, and Generic regularization. The performance of semi-supervised learning based on the MixMatch algorithm was compared with that of supervised learning using defect image data from the metal casting process. For the experiments, the ratio of labeled data was adjusted to 5%, 10%, 25%, and 50% of the total data. At a labeled data ratio of 5%, semi-supervised learning achieved a classification accuracy of 90.19%, outperforming supervised learning by approximately 22%p. At a 10% ratio, it surpassed supervised learning by around 8%p, achieving a 92.89% accuracy. These results demonstrate that semi-supervised learning can achieve significant outcomes even with a very limited amount of labeled data, suggesting its invaluable application in real-world research and industrial settings where labeled data is limited.
Abstract
Recently, there has been an increasing attempt to replace defect detection inspections in the manufacturing industry using deep learning techniques. However, obtaining substantial high-quality labeled data to enhance the performance of deep learning models entails economic and temporal constraints. As a solution for this problem, semi-supervised learning, using a limited amount of labeled data, has been gaining traction. This study assesses the effectiveness of semi-supervised learning in the defect detection process of manufacturing using the MixMatch algorithm. The MixMatch algorithm incorporates three dominant paradigms in the semi-supervised field: Consistency regularization, Entropy minimization, and Generic regularization. The performance of semi-supervised learning based on the MixMatch algorithm was compared with that of supervised learning using defect image data from the metal casting process. For the experiments, the ratio of labeled data was adjusted to 5%, 10%, 25%, and 50% of the total data. At a labeled data ratio of 5%, semi-supervised learning achieved a classification accuracy of 90.19%, outperforming supervised learning by approximately 22%p. At a 10% ratio, it surpassed supervised learning by around 8%p, achieving a 92.89% accuracy. These results demonstrate that semi-supervised learning can achieve significant outcomes even with a very limited amount of labeled data, suggesting its invaluable application in real-world research and industrial settings where labeled data is limited.
참고문헌 (21)
- Artificial Intelligence Index Report 2021, , McKinsey &Company, Stanford University, ~ (2021)
- Berthelot, D., "Mixmatch : A Holistic Approach to Semi-supervised Learning", Advances in Neural Information Processing Systems, 545, pp.5049~5059 (2019)
- "Casting Product Image Data for Quality Inspection"[URL]
- 전찬준, "수도 레이블을 활용한 준지도 학습 기반의 도로노면 파손 탐지", 한국ITS학회 논문지, 18(4), pp.71~79 (2019), 한국ITS학회[DOI]
- Dexin Gao, "NOVEL SEMI-SUPERVISED FAULT DIAGNOSIS METHOD COMBINING TRI-TRAINING AND DEEP BELIEF NETWORK FOR CHARGING EQUIPMENT OF ELECTRIC VEHICLE", International Journal of Automotive Technology, 23(6), pp.1727~1737 (2022), 한국자동차공학회
- Guo, C., "On calibration of modern neural networks", Interna-tional Conference on Machine Learning, PMLR, pp.1321~1330 (2017)
- Grandvalet, Y., "Semi-supervised Learning by Entropy Minimization", Advances in Neural Infor-mation Processing Systems, 17, pp.529~536 (2004)
- 김인기, "제한된 레이블 데이터에 대한 Mean Teacher 반지도학습을 활용한 견고한 포트홀 이미지 분류", 한국컴퓨터정보학회논문지, 28(5), pp.17~28 (2023), 한국컴퓨터정보학회[DOI]
- Kingma, D. P., "Semi-supervised Learning with Deep Generative Models", Advances in Neural Information Processing Systems, 268, pp.3581~3589 (2014)
- Kostopoulos, G., "Semi-supervised Regression : A Recent Review", Journal of Intelligent & Fuzzy Systems, 35(2), pp.1483~1500 (2018)
- Laine, S., "Temporal Ensembling for Semi-Supervised Learning", arXiv preprint arXiv, 2016, 1610.02242, ~ (2016)
- 이현로, "준지도 학습 모델을 사용한 차량 내부 네트워크에서의 이상 징후 탐지", 한국산학기술학회논문지, 24(2), pp.11~20 (2023), 한국산학기술학회[DOI]
- Lee, S. W., "Machine Vision-based Quality Inspection Method and System Utilizing Deep Learning in Manufacturing Process, 10-2017-0162907", ~ (2019)
- 이용호, "준지도학습 기반 반도체 공정 이상 상태 감지 및 분류", 반도체디스플레이기술학회지, 19(4), pp.121~125 (2020), 한국반도체디스플레이기술학회
- Quali, Y., "An Overview of Deep Semi-supervised Learning", arXiv preprint arXiv, 2020, 2006.05278, ~ (2020)
- Shim, S.B, "Adversarial Learning for Underground Structure Concrete Crack Detection Based on Semi-supervised Semantic Segmentation", Journal of Korean Tunnelling and Underground Space Association, 25(2), pp.515~528
- 신요섭, "AMFM : 금속 외관 결함의 준지도 분류에 특화된 AugMix 기반 FixMatch 알고리즘", 한국지능시스템학회 논문지, 32(5), pp.424~431 (2022), 한국지능시스템학회[DOI]
- Sun, S., "Anomaly Detection of Electricity Load Data Based on MixMatch", International Conference on Computer Engineering and Artificial Intelligence, pp.53~56 (2021)
- Yang, X., "A Survey on Deep Semi-supervised Learning", IEEE Transactions on Knowledge and Data Engineering, 35(9), pp.8934~8954 (2022)
- Zagoruyko, S., "Wide Residual Networks", arXiv preprint arXiv, 2016, 1605.07146, ~ (2016)
- Zhang, H., "Mixup: Beyond Empirical Risk Minimization", arXiv preprint arXiv, 2017, 1710.09412, ~ (2017)