Loading...
|
Please use this identifier to cite or link to this item:
https://nccur.lib.nccu.edu.tw/handle/140.119/158491
|
| Title: | 深度學習運用於紙本檔案修復影像多標籤分類之研究
Research on the Application of Deep Learning for Multi-label Classification of Restored Paper Archive Images |
| Authors: | 李維軒
LI, WEI-HSUAN |
| Contributors: | 林巧敏
Lin, Chiao-Min
李維軒
LI, WEI-HSUAN |
| Keywords: | 紙質檔案
檔案分類
多標籤分類
深度學習
Paper archives
Archive classification
Multi-label classification
Deep learning |
| Date: | 2025 |
| Issue Date: | 2025-08-04 14:03:57 (UTC+8) |
| Abstract: | 紙質檔案作為保存歷史記憶的重要載體，隨著時間演變，常因材質老化、環境變遷或人為操作而產生各種損壞。面對龐大的修復需求，傳統以人工判讀為主的處理方式，往往耗時且受主觀經驗限制，修復資源亦難以有效分配。在數位典藏與人工智慧技術快速發展的背景下，如何導入自動化技術協助檔案受損辨識與修復判斷，成為當前檔案修復實務與研究關注之課題。
本研究旨在應用深度學習技術建構紙本檔案受損分類系統，提升修復作業之效率與準確性。研究以《羅家倫》檔案影像為資料來源，針對十種常見受損類型（如變色泛黃、黴斑、皺褶痕等），建立多標籤分類架構，並實作DenseNet與Vision Transformer（ViT）兩種神經網路模型，搭配Binary Cross-Entropy、Focal Loss與Asymmetric Loss等損失函數進行訓練與交叉驗證，評估不同組合之分類效能。
本研究結果顯示，DenseNet結合Focal Loss於Precision與F1-score表現最佳，Asymmetric Loss則於Recall指標具優勢，顯示可依實務需求選擇適合之模型架構與損失設計；另對於樣本數量較少或特徵模糊之受損類別，預測效果則較不理想，反映標註一致性與資料量為影響模型準確性之關鍵因素。
為促進模型落實應用，研究亦建置網頁系統整合模型預測與結合知識庫生成建議說明，提供使用者進行影像上傳、受損判讀與修復決策參考之平台。整體而言，本研究證實深度學習技術可有效應用於紙本檔案受損分類，未來可持續透過資料集擴充與模型微調優化分類效能，並建置標註與回饋機制，發展智慧化之檔案修復輔助系統。
Paper-based archives serve as crucial carriers of historical memory. Over time, these materials are prone to various forms of deterioration due to material aging, environmental fluctuations, and human handling. In light of the substantial demand for restoration, traditional methods relying primarily on manual assessment are often time-consuming and subject to individual interpretation, resulting in inconsistent prioritization and inefficient allocation of restoration resources. Against the backdrop of rapid advancements in digital preservation and artificial intelligence technologies, the integration of automated approaches to support damage identification and restoration decision-making has emerged as a pressing issue within archival studies and practice.
This study aims to develop an automated classification system for paper-based archival damage using deep learning techniques, thereby enhancing the efficiency and objectivity of restoration workflows. The research utilizes a dataset comprising 1,149 annotated images from the Lo Chia-lun archive, encompassing ten common damage categories such as discoloration, mildew, and fold marks. A multi-label classification framework was implemented using two deep neural network architectures, DenseNet and Vision Transformer (ViT), in conjunction with three loss functions—Binary Cross-Entropy, Focal Loss, and Asymmetric Loss. These models were trained and evaluated through five-fold cross-validation to compare their classification performance across different configurations.
Experimental results indicate that the DenseNet model combined with Focal Loss achieved superior performance in terms of precision and F1-score, while Asymmetric Loss yielded higher recall, suggesting the selection of model-loss function combinations should be informed by specific practical requirements. Lower prediction accuracy was observed in categories with fewer samples or less distinguishable features, underscoring the importance of consistent labeling and sufficient data volume in multi-label learning tasks.
To facilitate practical deployment, a web-based system was developed, integrating model inference with a knowledge-based suggestion module. The platform allows users to upload images, receive automated damage assessments, and access corresponding restoration recommendations. Overall, the study demonstrates the applicability of deep learning to archival damage classification and provides a foundation for the development of intelligent restoration support systems through future efforts in dataset expansion, model fine-tuning, and the incorporation of expert feedback mechanisms. |
| Reference: | 一、中文文獻
Lien, Larry（2024）。數位轉型是什麼？成功轉型 3 階段與 4 種做法 | 政府企業補助資訊分享【2024】。https://www.hububble.co/blog/digitaltransformation#%E6%95%B8%E4%BD%8D%E8%BD%89%E5%9E%8B%E4%B8%89%E9%9A%8E%E6%AE%B5
何宣儀（2020）。應用人工智慧技術於圖書館紙本採購（未出版之碩士論文）。國立臺北科技大學資訊與財金管理系碩士班。
吳承恩、林巧敏（2023）。檔案內容主題自動分類及其成效評估之研究。檔案半年刊，22（2），34-53。
林巧敏（2022）。國立政治大學典藏中國國民黨檔案之整理與檢索服務。漢學研究通訊，41（2），33-39。
林素甘、楊美華、柯皓仁（2008）。數位化發展對檔案典藏與保存之影響。臺灣圖書館管理季刊，4（2），67-68。
政府出版品資訊網（2020）。【檔案局解密03】解密檔案修復：職人心．達人手藝．匠人魂。政府出版品資訊網。https://doi.org/10.6916/STPIRP.2019-06-20
洪敦明（2024）。AI 智慧館員：探索國立公共資訊圖書館「智慧服務」的演進之旅 [AI Librarians: Exploring the Evolutionary Journey of 'Smart Services' at the National Public Information Library]。臺北市立圖書館館訊，37（3），43-53。
國立公共資訊圖書館（2019）。讀者服務。《國立公共資訊圖書館 107 年年報》， 31- 33。
國立政治大學圖書館（2024）。羅家倫文庫。https://da.lib.nccu.edu.tw/sp-1.html
張鴻銘（2006）。現代化檔案保存維護與應用。海峽兩岸檔案暨微縮學術交流會論文集，15-28。
教育部智慧博物館（2017）。在圖書館上班的智能機器人。https://moeimo2016.blogspot.com/2017/07/blog-post_0.html
陳淑美、陳郁琳（2022）。國家檔案黏著劑劣化調查及動物膠移除之可行性初探。檔案半年刊，21（1），4-27。
陳靜儀、陳秋枝、賀語宸、李昕頤（2019）。國家檔案自動化管理之研究。檔案半年刊，18（2），72-89。
曾淑賢、廖箴 （2019）。圖書館通過對話實現變革-赴希臘雅典出席 [2019 年國際圖書館協會聯盟 (IFLA) 第 85 屆年會]。國家圖書館館訊，162，1-6。
黃元鶴（2022）。人工智慧技術應用與圖書館服務。臺灣出版與閱讀（111年第4），12-18。
臺灣藝術與建築索引典（2024）。修復[300053742]。https://aat.teldap.tw/AATFullDisplay/300053742
蔣佳蓉、戴芳伶、許尹馨（2022）。OCR 與機器學習在檔案內容辨識之初探。檔案半年刊，21（2），64-87。
檔案管理局（2002）。檔案管理名詞彙編。檔案管理局。
檔案管理局（2025）。國家發展委員會檔案管理局典藏國家檔案主要類別之內容大要。https://www.archives.gov.tw/tw/arctw/504.html
簡笙簧（1997）。紙質檔案破損修復概論。檔案與微縮，（45），13-28。
聶曼影（1993）。FMC紙張去酸加固法綜述。檔案學通訊，5。
二、英文文獻
Aggarwal, C. C., & Aggarwal, C. C. (2015). Data classification. Data classification (pp. 285-344). Springer International Publishing.
Barlow, H. B. (1989). Unsupervised learning. Neural Computation, 1(3), 295-311.
Bengio, Y., Simard, P., & Frasconi, P. (1994). Learning long-term dependencies with gradient descent is difficult. IEEE Transactions on Neural Networks, 5(2), 157-166.
Blakemore, E. (2016). High Tech Shelf Help: Singapore's Library Robot. https://www.libraryjournal.com/story/high-tech-shelf-help-singapores-library-robot
Blanke, T., & Wilson, J. (2017). Identifying epochs in text archives. 2017 IEEE International Conference on Big Data (Big Data).
Boureau, Y.-L., Ponce, J., & LeCun, Y. (2010). A theoretical analysis of feature pooling in visual recognition. Proceedings of the 27th International Conference on Machine Learning (ICML-10).
Brygfjeld, S. A., Wetjen, F., & Walsøe, A. (2017). Machine learning for production of Dewey Decimal. Proceedings of the International Federation of Library Associations and Institutions (IFLA).
City of Chicago. (2024). Chicago data portal. https://data.cityofchicago.org
Colavizza, G., Blanke, T., Jeurgens, C., & Noordegraaf, J. (2021). Archives and AI: An overview of current debates and future perspectives. ACM Journal on Computing and Cultural Heritage (JOCCH, 15(1), 1-15.
Cunningham, P., Cord, M., & Delany, S. J. (2008). Supervised learning. Machine learning techniques for multimedia: Case studies on organization and retrieval (pp. 21-49). Berlin, Heidelberg: Springer Berlin Heidelberg.
Dosovitskiy, A. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929.
Forde, H., & Rhys-Lewis, J. (2013). Preserving archives. London: Facet Publishing.
Fukushima, K. (1980). Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biological Cybernetics, 36, 193–202.
Han, K., Wang, Y., Chen, H., Chen, X., Guo, J., Liu, Z., ... & Tao, D. (2022). A survey on vision transformer. IEEE transactions on pattern analysis and machine intelligence, 45(1), 87-110.
Hearst, M. A., Dumais, S. T., Osuna, E., Platt, J., & Schölkopf, B. (1998). Support vector machines. IEEE Intelligent Systems and Their Applications, 13(4), 18-28.
Hengchen, S., Coeckelbergs, M., Van Hooland, S., Verborgh, R., & Steiner, T. (2016). Exploring archives with probabilistic models: Topic modelling for the valorisation of digitised archives of the European Commission. 2016 IEEE International Conference on Big Data (Big Data) (pp. 3447–3453).
Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, 313(5786), 504-507.
Hinton, G. E., Osindero, S., & Teh, Y. W. (2006). A fast learning algorithm for deep belief nets. Neural Computation, 18(7), 1527-1554.
Hochreiter, S. (1998). The vanishing gradient problem during learning recurrent neural nets and problem solutions. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 6(2), 107-116.
Hossin, M., & Sulaiman, M. N. (2015). A review on evaluation metrics for data classification evaluations. International Journal of Data Mining & Knowledge Management Process, 5(2), 1-11.
Hsu, M. W., & Lin, C. M. (2018). The use and restoration of papers for archival preservation: Case study of National Taiwan Library. Journal of Literature and Art Studies, 8(6), 977-988. https://doi.org/10.17265/2159-5836/2018.06.018
Huang, G., Chen, D., Li, T., Wu, F., Van Der Maaten, L., & Weinberger, K. Q. (2017). Multi-scale dense convolutional networks for efficient prediction. arXiv preprint arXiv:1703.09844, 2(2).
Hutchinson, T. (2020). Natural language processing and machine learning as practical toolsets for archival processing. Records Management Journal, 30(2), 155-174.
Ilievski, A., Zdraveski, V., & Gusev, M. (2018). How CUDA powers the machine learning revolution. Proceedings of the 26th Telecommunications Forum, TELFOR 2018 (pp. 420–425).
Jain, S. M. (2022). Introduction to transformers for NLP. With the Hugging Face Library and Models to Solve Problems.
Jordan, M. I., & Mitchell, T. M. (2015). Machine learning: Trends, perspectives, and prospects. Science, 349(6245), 255-260.
Kaelbling, L. P., Littman, M. L., & Moore, A. W. (1996). Reinforcement learning: A survey. Journal of Artificial Intelligence Research, 4, 237-285.
Karpathy, A. (2020, July 23). PyTorch at Tesla [Video]. YouTube. https://www.youtube.com/watch?v=oBklltKXtDE
Kingma, D. P., & Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 25. https://doi.org/10.1145/3065386
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.
Li, Z., Liu, F., Yang, W., Peng, S., & Zhou, J. (2021). A survey of convolutional neural networks: Analysis, applications, and prospects. IEEE Transactions on Neural Networks and Learning Systems, 33(12), 6999-7019.
Lin, T. Y., Goyal, P., Girshick, R., He, K., & Dollár, P. (2017). Focal loss for dense object detection. In Proceedings of the IEEE International Conference on Computer Vision (pp. 2980–2988).
Mahesh, B. (2020). Machine learning algorithms—a review. International Journal of Science and Research (IJSR), 9(1), 381-386.
Malik, D., & Munjal, G. (2021). Reviewing classification methods on health care. Intelligent Healthcare: Applications of AI in eHealth (pp. 127-142). Cham: Springer International Publishing.
Moss, M. S., & Gollins, T. J. (2017). Our digital legacy: An archival perspective. In Michael Moss on Archives (pp. 160-184). Routledge.
Paszke, A., Gross, S., Massa, F., et al. (2019). PyTorch: An Imperative Style, High-Performance Deep Learning Library. Advances in Neural Information Processing Systems (NeurIPS 2019), 32.
Raina, R., Madhavan, A., & Ng, A. Y. (2009). Large-scale deep unsupervised learning using graphics processors. Proceedings of the 26th Annual International Conference on Machine Learning. https://doi.org/10.1145/1553374.1553486
Ranade, S. (2016). Traces through time: A probabilistic approach to connected archival data. 2016 IEEE International Conference on Big Data (Big Data) (pp. 2316–2324).
Reddy, Y., Viswanath, P., & Reddy, B. E. (2018). Semi-supervised learning: A brief review. International Journal of Engineering & Technology, 7(1.8), 81.
Ridnik, T., Ben-Baruch, E., Zamir, N., Noy, A., Friedman, I., Protter, M., & Zelnik-Manor, L. (2021). Asymmetric loss for multi-label classification. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 82–91).
Rolan, G., Humphries, G., Jeffrey, L., Samaras, E., Antsoupova, T., & Stuart, K. (2019). More human than human? Artificial intelligence in the archive. Archives and Manuscripts, 47(2), 179-203.
Ruby, U., & Yendapalli, V. (2020). Binary cross entropy with deep learning technique for image classification. International Journal of Advanced Trends in Computer Science and Engineering, 9(10).
Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323, 533–536.
Sødring, T., Reinholdtsen, P., & Massey, D. (2020). A record-keeping approach to managing IoT-data for government agencies. Records Management Journal, 30(2), 221-239.
Sonmez, C., Özgür, A., & Yörük, E. (2016). Towards building a political protest database to explain changes in the welfare state. Proceedings of the 10th SIGHUM Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities (pp. 95–104).
Strong, A. (2016). Applications of artificial intelligence & associated technologies. Science [ETEBMS-2016], 5(6), 64-67.
Sublette, J. R. (1973). The Dartmouth conference: Its reports and results. College English, 35(3), 348-357.
The American Institute for Conservation of Historic & Artistic Works. (2024). Definitions of conservation terminology. Cultural Heritage. https://www.culturalheritage.org/about-conservation/what-is-conservation/specialties-in-conservation
The News Lens. (2017). McKinsey to Taiwan’s Execs: Digitalize or Die. https://international.thenewslens.com/article/79798
Theimer, K. (2018). It’s the end of the archival profession as we know it, and I feel fine. In C. Brown (Ed.), Archival Futures (pp. 1–18). Facet.
Tsoumakas, G., & Katakis, I. (2008). Multi-label classification: An overview. Data warehousing and mining: Concepts, methodologies, tools, and applications (pp. 64-74).
Turing, A. M. (1980). Computing machinery and intelligence. Creative Computing, 6(1), 44-53.
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., ... & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30.
Vellino, A., & Alberts, I. (2016). Assisting the appraisal of e-mail records with automatic classification. Records Management Journal, 26(3), 293-313.
Wang, M., Lu, S., Zhu, D., Lin, J., & Wang, Z. (2018). A high-speed and low-complexity architecture for softmax function in deep learning. 2018 IEEE Asia Pacific Conference on Circuits and Systems (APCCAS). 10.1109/APCCAS.2018.8605654
Wu, T., He, S., Liu, J., Sun, S., Liu, K., Han, Q. L., & Tang, Y. (2023). A brief overview of ChatGPT: The history, status quo and potential future development. IEEE/CAA Journal of Automatica Sinica, 10(5), 1122-1136.
Yu, Y., Pedrycz, W., & Miao, D. (2014). Multi-label classification by exploiting label correlations. Expert Systems with Applications, 41(6), 2989-3004.
Yusof, N. N., Mohamed, A., & Abdul-Rahman, S. (2015). Reviewing classification approaches in sentiment analysis. Soft Computing in Data Science: First International Conference, SCDS 2015, Putrajaya, Malaysia, September 2-3, 2015, Proceedings (Vol. 1).
Zeiler, M. (2014). Visualizing and understanding convolutional networks. European Conference on Computer Vision. arXiv preprint arXiv:1311.2901.
Zhang, M.-L., & Zhou, Z.-H. (2013). A review on multi-label learning algorithms. IEEE Transactions on Knowledge and Data Engineering, 26(8), 1819-1837. |
| Description: | 碩士
國立政治大學
圖書資訊與檔案學研究所
112155020 |
| Source URI: | http://thesis.lib.nccu.edu.tw/record/#G0112155020 |
| Data Type: | thesis |
| Appears in Collections: | [圖書資訊與檔案學研究所] 學位論文
|
Files in This Item:
| File |
Description |
Size | Format | |
|---|
| 502001.pdf | | 6198Kb | Adobe PDF | 0 | View/Open |
|
All items in 政大典藏 are protected by copyright, with all rights reserved.
|
