한국정보시스템학회지:정보시스템연구 (The Journal of Information Systems)

  • 제32권4호
  • /
  • Pages.149-162
  • /
  • 2023
  • /
  • 1229-8476(pISSN)
  • /
  • 2733-8770(eISSN)
한국정보시스템학회 (Korea Association of Information Systems)
권호 표지
DOI QR코드

DOI QR Code

직물 이미지 결함 탐지를 위한 딥러닝 기술 연구: 트랜스포머 기반 이미지 세그멘테이션 모델 실험

Deep Learning Models for Fabric Image Defect Detection: Experiments with Transformer-based Image Segmentation Models

  • 투고 : 2023.11.07
  • 심사 : 2023.12.17
  • 발행 : 2023.12.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

Purpose In the textile industry, fabric defects significantly impact product quality and consumer satisfaction. This research seeks to enhance defect detection by developing a transformer-based deep learning image segmentation model for learning high-dimensional image features, overcoming the limitations of traditional image classification methods. Design/methodology/approach This study utilizes the ZJU-Leaper dataset to develop a model for detecting defects in fabrics. The ZJU-Leaper dataset includes defects such as presses, stains, warps, and scratches across various fabric patterns. The dataset was built using the defect labeling and image files from ZJU-Leaper, and experiments were conducted with deep learning image segmentation models including Deeplabv3, SegformerB0, SegformerB1, and Dinov2. Findings The experimental results of this study indicate that the SegformerB1 model achieved the highest performance with an mIOU of 83.61% and a Pixel F1 Score of 81.84%. The SegformerB1 model excelled in sensitivity for detecting fabric defect areas compared to other models. Detailed analysis of its inferences showed accurate predictions of diverse defects, such as stains and fine scratches, within intricated fabric designs.

키워드

과제정보

이 논문은 2023년도 산업통상자원부 산업혁신기반구축사업 재원으로 수행된 연구임(P0014711).

참고문헌

  1. 김선아, 김정원, 원동연, 최예림, "무슬림 관광객 증대를 위한 머신러닝 기반의 할랄푸드 분류 프레임워크," 정보시스템연구, 제26권, 제3호, 2017, pp. 273-293.
  2. 이동훈, 김태형, "머신러닝 기법을 활용한 대졸 구직자 취업 예측모델에 관한 연구," 정보시스템연구, 제29권, 제2호, 2020, pp. 287-306.
  3. 한국섬유경제, "다이텍硏, '섬유소재 빅데이터 통합 지원센터' 섬유소재기업-바이어 'ON'", 김진일 기고, 2022.06.27.
  4. Almeida, T., Moutinho, F., and Matos Carvalho, J. P., "Deep Learning for Fabric Defect Detection with False Negative Reduction," IEEE, Vol. 9, 2021, pp. 81936-81945. https://doi.org/10.1109/ACCESS.2021.3086028
  5. Apple, "MobileViT + DeepLabV3 (small-sized model)", https://huggingface.co/apple/deeplabv3-mobilevit-small, 2021.
  6. Chen, L., Papandreou, G., Schroff, F., and Adam, H., "Rethinking Atrous Convolution for Semantic Image Segmentation," arXiv preprint arXiv:1706.05587, 2017.
  7. Cheng, B., Schwing, A., Kirillov, and A., "Per-Pixel Classification Is Not All You Need for Semantic Segmentation," Advances in Neural Information Processing Systems, Vol. 34, 2021, pp. 17864-17875.
  8. Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., and Unterthiner, T., "Transformers for Image Recognition at Scale," arXiv preprint arXiv: 2010.11929, 2020.
  9. Huang, Y., Jing, J., and Wang, Z., "Fabric Defect Segmentation Method Based on Deep Learning," IEEE Transactions on Instrumentation and Measurement, Vol. 70, 2021, Article 1000315.
  10. LeCun, Y., Bottou, L., Bengio, Y., and Haffner, P., "Gradient-Based Learning Applied to Document Recognition," Proceedings of the IEEE, Vol. 86, No. 11, 1998, pp. 2278-2324. https://doi.org/10.1109/5.726791
  11. Liu, Q., Wang, C., Li, Y., Gao, M., and Li, J., "A Fabric Defect Detection Method Based on Deep Learning," IEEE, Vol. 10, 2022, pp. 4284-4296. https://doi.org/10.1109/ACCESS.2021.3140118
  12. Lu, D., Weng, Q., and "A Survey of Image Classification Methods and Techniques for Improving Classification Performance," International Journal of Remote Sensing, Vol. 28, No. 5, 2007, pp. 823-870. https://doi.org/10.1080/01431160600746456
  13. Mehta, S., and Rastegari, M., "MobileViT: Light-Weight, General- Purpose, and Mobile-Friendly Vision Transformer," arXiv preprint arXiv: 2110.02178, 2021.
  14. Meta, "Vision Transformer (small-sized model) trained using DINOv2", https://huggingface.co/facebook/dinov2-small, 2023.
  15. Minaee, S., Boykov, Y., Porikli, F., Plaza, A., Kehtarnavaz, N., and Terzopoulos, D., "Image Segmentation Using Deep Learning: A Survey," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 44, No. 7, 2021, pp. 3523-3542. https://doi.org/10.1109/TPAMI.2021.3059968
  16. Nvidia, "SegFormer (b0-sized) model fine-tuned on ADE20k", https://huggingface.co/nvidia/segformer-b0-finetuned-ade-512-512, 2021.
  17. Oquab, M., Darcet, T., Moutakanni, T., Vo, H., Szafraniec, M., Khalidov, V., and Fernandez, P et al., "DinoV2: Learning Robust Visual Features without Supervision," arXiv preprint arXiv:2304.07193, 2023.
  18. Shahrabadi, S., Castilla, Y., Guevara, M., Magalhaes, L. G., Gonzalez, D., and Adao, T., "Defect Detection in the Textile Industry Using Image-Based Machine Learning Methods: A Brief Review," the Journal of Physics: Conference Series, 2022.
  19. Voronin, V., Sizyakin, R., Zhdanova, M., Semenishchev, E., Bezuglov, D., and Zelemskii, A., "Automated Visual Inspection of Fabric Image Using Deep Learning Approach for Defect Detection," the Automated Visual Inspection and Machine Vision IV, 2021.
  20. Wei, B., Xu, B., Hao, K., and Gao, L., "Textile Defect Detection Using Multilevel and Attentional Deep Learning Network (MLMA-Net)," Textile Research Journal, Vol. 92, No. 19-20, 2022, pp. 3462-3477. https://doi.org/10.1177/00405175211073773
  21. Xie, E., Wang, W., Yu, Z., Anandkumar, A., Alvarez, J. M., and Luo, P., "SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers," Advances in Neural Information Processing Systems, Vol. 34, 2021, pp. 12077-12090.
  22. Zhang, C., Feng, S., Wang, X., and Wang, Y., "ZJU-Leaper: A Benchmark Dataset for Fabric Defect Detection and a Comparative Study," IEEE Transactions on Artificial Intelligence, Vol. 1, No. 3, 2020, pp. 219-232. https://doi.org/10.1109/TAI.2021.3057027
  23. Zhao, Z.-Q., Zheng, P., Xu, S., and Wu, X., "Object Detection with Deep Learning: A Review," IEEE Transactions on Neural Networks and Learning Systems, Vol. 30, No. 11, 2019, pp. 3212-3232. https://doi.org/10.1109/TNNLS.2018.2876865