Publications-Theses
Article View/Open
Publication Export
-
題名 基於深度學習框架之電器火災電線金相識別與應用
Metallographic Analysis of Electric Wires in Fire Accidents Using Deep Learning Approaches作者 彭建凱
Peng, Chien-Kai貢獻者 廖文宏
Liao, Wen-Hung
彭建凱
Peng, Chien-Kai關鍵詞 深度學習
圖像分類
遷移學習
資料增強
模型可解釋化
導線熔痕分析
Deep learning
Image classification
Transfer learning
Data Augmentation
Model interpretability
Metallographic Analysis日期 2021 上傳時間 2-Mar-2021 14:56:59 (UTC+8) 摘要 本論文試圖探究如何在資料集高度不平衡且稀少的情況下，利用深度學習之方法，將火災現場所取得之巨觀以及微觀之導線熔痕進行分類，並以Grad-CAM的方法分析深度學習模型所學習之特徵。\n本研究所使用的方法，將使用深度學習中遷移學習之概念訓練模型，同時透過資料增強的方法，擴充並平衡資料集之分布，以提高熔痕識別之效能。經過資料增強、資料清理、模型優化與參數調校後，最佳實驗結果得出巨觀通電痕 F1-Score 89.22%、巨觀熱熔痕 F1-Score 80.85%、微觀通電痕 F1-Score 79.46%、微觀熱熔痕 F1-Score 81.90%，並同步建置可用實務上之應用程式原型，達到輔助現場判決與進一步蒐集資料之目的，也期許這樣的成果可提升實務之效率，以提供相關政策制定之參考。\n未來希望能以此為基礎，探討更進一步優化導線金相之識別分法，並投入到更多的應用當中，持續改善實務之工作流程。
The objective of this thesis aims to classify the wire melting marks from fire scenes based on deep learning approaches when the data set is imbalanced and only a limited amount of data is available. The correctness of the results is verified through the Grad-CAM method.\nThis thesis employs the concept of transfer learning to train models, and balance the distribution of the data set through the method of data augmentation, so as to improve the efficiency of melting mark recognition. After data augmentation, data cleaning, model optimization and parameter fine-tuning, the best experimental results in terms of F1 are: 89.22% for macro electricity mark, 80.85% for macro heat-melting mark, 79.46% for micro electricity mark, and 81.90% for micro heat-melting mark. An application prototype has been built to assist on-site recognition and further data collection. It is hoped that the results can enhance the performance and provide references for policies making.\nIn addition to laying the foundation for further optimizing the wire melting marks identification method, this thesis also improves task efficiency and government`s work flow.參考文獻 [1] 中華民國內政部消防署全球資訊網 火災統計 https://www.nfa.gov.tw/cht/index.php?code=list&ids=220\n[2] 中華民國內政部消防署全球資訊網 修正「火災調查鑑定標準作業程序」、「火災原因調查鑑定書製作規定」、「火災原因調查鑑定書分級列管實施規定」之名稱及規定 https://www.nfa.gov.tw/cht/index.php?code=list&flag=detail&ids=23&article_id=343\n[3] Johnson, J. M., & Khoshgoftaar, T. M. (2019). Survey on deep learning with class imbalance. Journal of Big Data, 6(1), 27.\n[4] Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357.\n[5] Elkan, C. (2001, August). The foundations of cost-sensitive learning. In International joint conference on artificial intelligence (Vol. 17, No. 1, pp. 973-978). Lawrence Erlbaum Associates Ltd.\n[6] Tan, C., Sun, F., Kong, T., Zhang, W., Yang, C., & Liu, C. (2018, October). A survey on deep transfer learning. In International conference on artificial neural networks (pp. 270-279). Springer, Cham.\n[7] Goodman, B., & Flaxman, S. (2017). European Union regulations on algorithmic decision-making and a “right to explanation”. AI magazine, 38(3), 50-57.\n[8] Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., & Torralba, A. (2016). Learning deep features for discriminative localization. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 2921-2929).\n[9] Lin, M., Chen, Q., & Yan, S. (2013). Network in network. arXiv preprint arXiv:1312.4400.\n[10] Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., & Batra, D. (2017). Grad-CAM: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision (pp. 618-626).\n[11] Springenberg, J. T., Dosovitskiy, A., Brox, T., & Riedmiller, M. (2014). Striving for simplicity: The all convolutional net. arXiv preprint arXiv:1412.6806.\n[12] Chattopadhay, A., Sarkar, A., Howlader, P., & Balasubramanian, V. N. (2018, March). Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks. In 2018 IEEE Winter Conference on Applications of Computer Vision (WACV) (pp. 839-847). IEEE.\n[13] ImageNet http://www.image-net.org/\n[14] Coco DataSet https://cocodataset.org/\n[15] Open Image DataSet https://storage.googleapis.com/openimages/web/index.html\n[16] Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. (2017, February). Inception-v4, inception-resnet and the impact of residual connections on learning. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 31, No. 1). 描述 碩士
國立政治大學
資訊科學系碩士在職專班
107971001資料來源 http://thesis.lib.nccu.edu.tw/record/#G0107971001 資料類型 thesis dc.contributor.advisor 廖文宏 zh_TW dc.contributor.advisor Liao, Wen-Hung en_US dc.contributor.author (Authors) 彭建凱 zh_TW dc.contributor.author (Authors) Peng, Chien-Kai en_US dc.creator (作者) 彭建凱 zh_TW dc.creator (作者) Peng, Chien-Kai en_US dc.date (日期) 2021 en_US dc.date.accessioned 2-Mar-2021 14:56:59 (UTC+8) - dc.date.available 2-Mar-2021 14:56:59 (UTC+8) - dc.date.issued (上傳時間) 2-Mar-2021 14:56:59 (UTC+8) - dc.identifier (Other Identifiers) G0107971001 en_US dc.identifier.uri (URI) https://ah.lib.nccu.edu.tw/item?item_id=153229 - dc.description (描述) 碩士 zh_TW dc.description (描述) 國立政治大學 zh_TW dc.description (描述) 資訊科學系碩士在職專班 zh_TW dc.description (描述) 107971001 zh_TW dc.description.abstract (摘要) 本論文試圖探究如何在資料集高度不平衡且稀少的情況下，利用深度學習之方法，將火災現場所取得之巨觀以及微觀之導線熔痕進行分類，並以Grad-CAM的方法分析深度學習模型所學習之特徵。\n本研究所使用的方法，將使用深度學習中遷移學習之概念訓練模型，同時透過資料增強的方法，擴充並平衡資料集之分布，以提高熔痕識別之效能。經過資料增強、資料清理、模型優化與參數調校後，最佳實驗結果得出巨觀通電痕 F1-Score 89.22%、巨觀熱熔痕 F1-Score 80.85%、微觀通電痕 F1-Score 79.46%、微觀熱熔痕 F1-Score 81.90%，並同步建置可用實務上之應用程式原型，達到輔助現場判決與進一步蒐集資料之目的，也期許這樣的成果可提升實務之效率，以提供相關政策制定之參考。\n未來希望能以此為基礎，探討更進一步優化導線金相之識別分法，並投入到更多的應用當中，持續改善實務之工作流程。 zh_TW dc.description.abstract (摘要) The objective of this thesis aims to classify the wire melting marks from fire scenes based on deep learning approaches when the data set is imbalanced and only a limited amount of data is available. The correctness of the results is verified through the Grad-CAM method.\nThis thesis employs the concept of transfer learning to train models, and balance the distribution of the data set through the method of data augmentation, so as to improve the efficiency of melting mark recognition. After data augmentation, data cleaning, model optimization and parameter fine-tuning, the best experimental results in terms of F1 are: 89.22% for macro electricity mark, 80.85% for macro heat-melting mark, 79.46% for micro electricity mark, and 81.90% for micro heat-melting mark. An application prototype has been built to assist on-site recognition and further data collection. It is hoped that the results can enhance the performance and provide references for policies making.\nIn addition to laying the foundation for further optimizing the wire melting marks identification method, this thesis also improves task efficiency and government`s work flow. en_US dc.description.tableofcontents 第一章\n1.1 研究動機 1\n1.2 需求分析 5\n1.3 研究貢獻 5\n1.4 論文架構 6\n第二章\n2.1 不平衡資料 8\n2.1.1 資料級(data-level)方法 9\n2.1.2 演算法級(algorithm-level)方法 11\n2.1.3 混合方法 11\n2.2 遷移學習 11\n2.3 交叉驗證 14\n2.3.1 Holdout 15\n2.3.2 K-Fold 15\n2.3.3 Leave-One-Out 16\n2.4 Grad-CAM 17\n第三章\n3.1 基本構想 23\n3.2 前期研究 24\n3.2.1 資料蒐集 24\n3.2.2 資料前處理 26\n3.2.3 遷移學習模型 27\n3.3 研究架構與設計 29\n3.3.1 問題陳述 29\n3.3.2 研究步驟 30\n3.4 目標設定 32\n第四章\n4.1 研究過程 33\n4.2 資料增強 34\n4.3 基線效能 35\n4.4 資料清理 38\n4.5 模型優化 46\n4.6 應用程式建置與操作流程 50\n4.7 訓練成果分析與探討 53\n第五章\n5.1 結論 57\n5.2 未來研究方向 57\n參考文獻 59 zh_TW dc.format.extent 8394257 bytes - dc.format.mimetype application/pdf - dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0107971001 en_US dc.subject (關鍵詞) 深度學習 zh_TW dc.subject (關鍵詞) 圖像分類 zh_TW dc.subject (關鍵詞) 遷移學習 zh_TW dc.subject (關鍵詞) 資料增強 zh_TW dc.subject (關鍵詞) 模型可解釋化 zh_TW dc.subject (關鍵詞) 導線熔痕分析 zh_TW dc.subject (關鍵詞) Deep learning en_US dc.subject (關鍵詞) Image classification en_US dc.subject (關鍵詞) Transfer learning en_US dc.subject (關鍵詞) Data Augmentation en_US dc.subject (關鍵詞) Model interpretability en_US dc.subject (關鍵詞) Metallographic Analysis en_US dc.title (題名) 基於深度學習框架之電器火災電線金相識別與應用 zh_TW dc.title (題名) Metallographic Analysis of Electric Wires in Fire Accidents Using Deep Learning Approaches en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) [1] 中華民國內政部消防署全球資訊網 火災統計 https://www.nfa.gov.tw/cht/index.php?code=list&ids=220\n[2] 中華民國內政部消防署全球資訊網 修正「火災調查鑑定標準作業程序」、「火災原因調查鑑定書製作規定」、「火災原因調查鑑定書分級列管實施規定」之名稱及規定 https://www.nfa.gov.tw/cht/index.php?code=list&flag=detail&ids=23&article_id=343\n[3] Johnson, J. M., & Khoshgoftaar, T. M. (2019). Survey on deep learning with class imbalance. Journal of Big Data, 6(1), 27.\n[4] Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357.\n[5] Elkan, C. (2001, August). The foundations of cost-sensitive learning. In International joint conference on artificial intelligence (Vol. 17, No. 1, pp. 973-978). Lawrence Erlbaum Associates Ltd.\n[6] Tan, C., Sun, F., Kong, T., Zhang, W., Yang, C., & Liu, C. (2018, October). A survey on deep transfer learning. In International conference on artificial neural networks (pp. 270-279). Springer, Cham.\n[7] Goodman, B., & Flaxman, S. (2017). European Union regulations on algorithmic decision-making and a “right to explanation”. AI magazine, 38(3), 50-57.\n[8] Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., & Torralba, A. (2016). Learning deep features for discriminative localization. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 2921-2929).\n[9] Lin, M., Chen, Q., & Yan, S. (2013). Network in network. arXiv preprint arXiv:1312.4400.\n[10] Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., & Batra, D. (2017). Grad-CAM: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision (pp. 618-626).\n[11] Springenberg, J. T., Dosovitskiy, A., Brox, T., & Riedmiller, M. (2014). Striving for simplicity: The all convolutional net. arXiv preprint arXiv:1412.6806.\n[12] Chattopadhay, A., Sarkar, A., Howlader, P., & Balasubramanian, V. N. (2018, March). Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks. In 2018 IEEE Winter Conference on Applications of Computer Vision (WACV) (pp. 839-847). IEEE.\n[13] ImageNet http://www.image-net.org/\n[14] Coco DataSet https://cocodataset.org/\n[15] Open Image DataSet https://storage.googleapis.com/openimages/web/index.html\n[16] Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. (2017, February). Inception-v4, inception-resnet and the impact of residual connections on learning. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 31, No. 1). zh_TW dc.identifier.doi (DOI) 10.6814/NCCU202100228 en_US
