學術產出-Theses
Article View/Open
Publication Export
-
題名 使用深度學習識別現代畫中的藝術風格-以台灣鄉土風格繪畫為例
Identifying Artistic Styles in Modern Painting Using Deep Learning- Taiwan's local style painting as an Example作者 陳亦霖
Chen, I-Lin貢獻者 彭彥璁
Peng, Yan-Tsung
陳亦霖
Chen, I-Lin關鍵詞 深度學習
圖像分類
台灣鄉土風格繪畫
資料擴增
Deep learning
Image classification
Taiwan's local style painting
Data augmentation日期 2024 上傳時間 1-Mar-2024 14:12:04 (UTC+8) 摘要 繪畫風格分類一直是一個活躍的研究領域，但對於國內繪畫風格透過深度學習分析的研究相當稀少，故本研究旨在透過機器學習協助現代藝術中有定義的國別風格主義分類與台灣鄉土繪畫風格一同進行分類探討，其中台灣鄉土繪畫樣本為本研究所收集從1860年代至1970年代這段時期內台灣知名前輩創作的本土鄉情藝術作品，以此做為台灣鄉土藝術風格繪畫的指標，期間台灣知名畫家有陳澄波、廖繼春、李梅樹、楊三郎、李石樵等人的繪畫作品，本研究稱為台灣鄉土風格繪畫，與WikiArt之現代藝術中可稱為國別的藝術主義風格一同進行監督式學習，並探討擴增資料集、訓練資料特性及分類器學習模效能之評分及人文歷史角度探討關聯。在本文中會利用常見的4種深度學習模型AlexNet、VGG19、GoogleNet、ResNet152，用以辨識和分類藝術風格，藝術風格分為美國寫實主義(American Realism)、日本主義(Japonism)、墨西哥壁畫運動(Muralism) 、印度空間畫(Indian Space painting)、巴洛克復興風格(Neo-baroque)、台灣鄉土藝術(Tw-Local) 等6種不同類型，透過資料集擴增處理及各模型選擇之實驗，最終以ResNet152模型分類準確率達到93%之表現，說明本研究定義的分類藝術風格可透過模型加以分類，並透過人文歷史角度探討深度學習模型所識別的繪畫特徵關聯做說明。
Painting style classification has always been an active research field, but research on domestic painting styles analyzed through deep learning is rare, This study aims to contribute to the classification of national mannerism in modern art by exploring the classification of Taiwanese vernacular painting style through machine learning. Among the aspects discussed, the samples of Taiwanese local paintings are local nostalgic works of art collected by the Institute from the 1860s to the 1970s, created by well-known predecessors in Taiwan, this can be taken as an indicator of Taiwan's local art style painting,during this period, well-known Taiwanese painters Chen Chengbo, Liao Jichun, Li Meishu, Yang Sanlang, Li Shiqiao and others produced their paintings. This study titled ‘Tw-Local style painting’ employs supervised learning to analyze what may be referred to as national artistic styles found in WikiArt's Modern Art.It also discuss the relationship between augmented data sets, training data characteristics and classifier learning model performance scores from a humanistic and historical perspective. In the text, we will use the four common deep learning models AlexNet, VGG19, GoogleNet, ResNet152. These models will be used to identify and classify artistic styles, which are divided into American Realism, Japonism, Muralism, Indian Space painting, and Neo-baroque, Taiwan local art (Tw-Local) and other 6 different types.The experiments involve data augmentation processing and selection of each model. Finally, using the ResNet152 model, the classification accuracy rate reached 93%. This result indicates that the categorical artistic styles defined in this study can be classified through the model and explains the correlation of painting features identified by the deep learning model from the perspective of humanities and history.參考文獻 [1] K. Richman-Abdou, "What is Modern Art? Exploring the Movements That Define the Groundbreaking Genre," MY MODERN MET, 2022. https://mymodernmet.com/what-is-modern-art-definition/. [2] C. Ives, "Japonisme," New York: The Metropolitan Museum of Art, 2004. http://www.metmuseum.org/toah/hd/jpon/hd_jpon.htm. [3] 陳曼華, 藝術與文化政治：戰後台灣藝術的主體形構, 國立交通大學, 2016. [4] "Wikiart," https://www.wikiart.org/. [5] "American Realism," https://reurl.cc/RyRbLZ. [6] "Japonism," https://reurl.cc/RyRbLZ. [7] "Muralism," https://reurl.cc/ka9Lmd. [8] "Indian Space painting," https://reurl.cc/x6eEdz. [9] "Neo-baroque," https://reurl.cc/NyOZze. [10] Y. LeCun., Y. Bengio.,G. Hinton., "Deep learning," Nature, no. 521(7553), p. 436–444, 2015. [11] 簡秀枝, 台灣前輩油畫家市場之研究—以陳澄波、廖繼春、李梅樹、楊三郎、李石樵之油畫市場行情為例, 國立臺灣師範大學, 2007. [12] 倪再沁, "西方美術．台灣製造-台灣現代美術的批判," 雄獅美術, no. 242, pp. 132-133, 1991. [13] 劉聖秋, 70年代台灣鄉土美術之研究, 國立屏東師範學院, 2002. [14] 陸蓉之, "台灣地區當代藝術本土風格語彙的衍變," in 中華民國美術思潮研討會論文集, 台北市立美術館, 1992. [15] “臺北市立美術館,” https://www.tfam.museum/index.aspx?ddlLang=zh-tw. [16] “楊三郎美術館,” https://yangsanlang.com.tw/collection/. [17] “國立台灣美術館,” https://ntmofa-collections.ntmofa.gov.tw/Default.aspx. [18] E. Ahmed., M. Marian., B. Liu., K. Diana., E. Mohamed., The Shape of Art History in the Eyes of the Machine, arXiv, 2018. [19] A. Krizhevsky., I. Sutskever., G. E. Hinton., ImageNet Classification with Deep Convolutional Neural Networks, ACM, 2012. [20] S. Kumar., A. Tyagi., T. Sahu., P. Shukla., A. Mittal., Indian Art Form Recognition Using Convolutional Neural Networks, IEEE, 2018. [21] D. Kvak., Leveraging Computer Vision Application in Visual Arts: A Case Study on the Use of Residual Neural Network to Classify and Analyze Baroque Paintings, arXiv, 2022. [22] V. H. Phung., E. J. Rhee., A High-Accuracy Model Average Ensemble of Convolutional Neural Networks for Classification of Cloud Image Patches on Small Datasets, Hanbat National University, 2019. [23] “維基百科： Overfiting說明,” https://reurl.cc/x6o2mN. [24] K. Simonyan., A. Zisserman., Very Deep Convolutional Networks for Large-Scale Image Recognition, arXiv, 2014. [25] C. Szegedy., W. Liu., Y. Jia,P. Sermanet., S. Reed.,D. Anguelov.,A. Rabinovich., Going Deeper with Convolutions, IEEE, 2015. [26] K. He.,X. Zhang.,S. Ren.,J. Sun., Deep Residual Learning for Image Recognition, IEEE, 2016. [27] 李謦伊, “卷積神經網絡 CNN 經典模型 — LeNet、AlexNet、VGG、NiN with Pytorch code,” Medium, 2020. https://reurl.cc/q06LpE. [28] C. Shorte.,T. M. Khoshgoftaar., A Survey on Image Data Augmentation for Deep Learning. Journal of Big Data, SpringerOpen, 2019. [29] V. d. Maaten., G. Hinton., "Visualizing data using t-SNE," Journal of machine learning research, vol. 9, no. 11, 2008. [30] G. Hinton., E. Geoffrey., Roweis., T. Sam, "Stochastic neighbor embedding.," Advances in neural information processing systems., pp. 857-864, 2002. [31] 謝東山, “臺灣鄉土美術的質與量,” 臺灣美術期刊, 編號 107, p. 6~8, 2017. [32] 葉國新, 傳統與創新--由鄉土藝術出發探討繪畫創作的表現, 國立臺灣師範大學, 2005. [33] R. R. Selvaraju.,M. Cogswell.,A. Das,R. Vedantam.,D. Parikh.,D. Batra., Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization, IEEE, 2017. [34] “維基百科：浮世檜,” https://reurl.cc/Nyngm5. [35] “維基百科：墨西哥壁畫運動,” https://reurl.cc/8NzRZo. [36] “維基百科：巴洛克藝術,” https://reurl.cc/r6o2Gb. [37] "台灣鄉土美術運動," 維基百科, https://reurl.cc/gaNz6R. [38] 許綾讌, 舊園情懷 ─ 台灣鄉土古厝水墨畫之研究, 國立臺灣藝術大學, 2010. 描述 碩士
國立政治大學
資訊科學系碩士在職專班
109971022資料來源 http://thesis.lib.nccu.edu.tw/record/#G0109971022 資料類型 thesis dc.contributor.advisor 彭彥璁 zh_TW dc.contributor.advisor Peng, Yan-Tsung en_US dc.contributor.author (Authors) 陳亦霖 zh_TW dc.contributor.author (Authors) Chen, I-Lin en_US dc.creator (作者) 陳亦霖 zh_TW dc.creator (作者) Chen, I-Lin en_US dc.date (日期) 2024 en_US dc.date.accessioned 1-Mar-2024 14:12:04 (UTC+8) - dc.date.available 1-Mar-2024 14:12:04 (UTC+8) - dc.date.issued (上傳時間) 1-Mar-2024 14:12:04 (UTC+8) - dc.identifier (Other Identifiers) G0109971022 en_US dc.identifier.uri (URI) https://nccur.lib.nccu.edu.tw/handle/140.119/150261 - 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) 109971022 zh_TW dc.description.abstract (摘要) 繪畫風格分類一直是一個活躍的研究領域，但對於國內繪畫風格透過深度學習分析的研究相當稀少，故本研究旨在透過機器學習協助現代藝術中有定義的國別風格主義分類與台灣鄉土繪畫風格一同進行分類探討，其中台灣鄉土繪畫樣本為本研究所收集從1860年代至1970年代這段時期內台灣知名前輩創作的本土鄉情藝術作品，以此做為台灣鄉土藝術風格繪畫的指標，期間台灣知名畫家有陳澄波、廖繼春、李梅樹、楊三郎、李石樵等人的繪畫作品，本研究稱為台灣鄉土風格繪畫，與WikiArt之現代藝術中可稱為國別的藝術主義風格一同進行監督式學習，並探討擴增資料集、訓練資料特性及分類器學習模效能之評分及人文歷史角度探討關聯。在本文中會利用常見的4種深度學習模型AlexNet、VGG19、GoogleNet、ResNet152，用以辨識和分類藝術風格，藝術風格分為美國寫實主義(American Realism)、日本主義(Japonism)、墨西哥壁畫運動(Muralism) 、印度空間畫(Indian Space painting)、巴洛克復興風格(Neo-baroque)、台灣鄉土藝術(Tw-Local) 等6種不同類型，透過資料集擴增處理及各模型選擇之實驗，最終以ResNet152模型分類準確率達到93%之表現，說明本研究定義的分類藝術風格可透過模型加以分類，並透過人文歷史角度探討深度學習模型所識別的繪畫特徵關聯做說明。 zh_TW dc.description.abstract (摘要) Painting style classification has always been an active research field, but research on domestic painting styles analyzed through deep learning is rare, This study aims to contribute to the classification of national mannerism in modern art by exploring the classification of Taiwanese vernacular painting style through machine learning. Among the aspects discussed, the samples of Taiwanese local paintings are local nostalgic works of art collected by the Institute from the 1860s to the 1970s, created by well-known predecessors in Taiwan, this can be taken as an indicator of Taiwan's local art style painting,during this period, well-known Taiwanese painters Chen Chengbo, Liao Jichun, Li Meishu, Yang Sanlang, Li Shiqiao and others produced their paintings. This study titled ‘Tw-Local style painting’ employs supervised learning to analyze what may be referred to as national artistic styles found in WikiArt's Modern Art.It also discuss the relationship between augmented data sets, training data characteristics and classifier learning model performance scores from a humanistic and historical perspective. In the text, we will use the four common deep learning models AlexNet, VGG19, GoogleNet, ResNet152. These models will be used to identify and classify artistic styles, which are divided into American Realism, Japonism, Muralism, Indian Space painting, and Neo-baroque, Taiwan local art (Tw-Local) and other 6 different types.The experiments involve data augmentation processing and selection of each model. Finally, using the ResNet152 model, the classification accuracy rate reached 93%. This result indicates that the categorical artistic styles defined in this study can be classified through the model and explains the correlation of painting features identified by the deep learning model from the perspective of humanities and history. en_US dc.description.tableofcontents 摘要 i Abstract ii 目錄 iv 圖目錄 vii 表目錄 ix 第一章 緒論 11 1.1 研究背景及動機 11 1.2 研究目的 12 1.3 論文貢獻 13 第二章 文獻探討 14 2.1 台灣鄉土風格藝術繪畫 14 2.2 深度學習中的繪畫風格分類相關文獻 16 2.3 類神經網路 18 2.4 深度學習 18 2.5 深度學習模型 20 2.5.1 AlexNet 21 2.5.2 VGG19 22 2.5.3 GoogLeNet 23 2.5.4 ResNet 24 2.6 圖像資料擴增 25 2.7 t-SNE 28 第三章 研究方法 30 3.1 研究方法 30 3.2 資料集 31 3.2.1 美國寫實主義 American Realism 32 3.2.2 日本主義Japonism 33 3.2.3 墨西哥壁畫運動風格 Muralism 34 3.2.4 印度空間畫Indian Space painting 35 3.2.5 巴洛克復興風格 Neo-baroque 36 3.2.6 台灣鄉土藝術風格繪畫 37 3.3 模型評估指標 40 3.3.1 混淆矩陣 40 3.3.2 準確率、精確率及召回率 41 3.3.3 K折交叉驗證 42 第四章 實驗結果及討論 43 4.1 資料處理 43 4.2 實驗模型及實驗環境 45 4.3交叉驗證實驗及結果 46 4.4 AlexNet實驗及結果 46 4.5 VGG19實驗及結果 48 4.6 GoogleNet實驗及結果 49 4.7 ResNet152實驗及結果 51 4.8各模型實驗結果比較 52 4.9 深度學習對於藝術風格分析 56 4.10 t-SNE資料分析 64 第五章 結論 69 參考文獻 71 zh_TW dc.format.extent 5452186 bytes - dc.format.mimetype application/pdf - dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0109971022 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) 台灣鄉土風格繪畫 zh_TW dc.subject (關鍵詞) 資料擴增 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) Taiwan's local style painting en_US dc.subject (關鍵詞) Data augmentation en_US dc.title (題名) 使用深度學習識別現代畫中的藝術風格-以台灣鄉土風格繪畫為例 zh_TW dc.title (題名) Identifying Artistic Styles in Modern Painting Using Deep Learning- Taiwan's local style painting as an Example en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) [1] K. Richman-Abdou, "What is Modern Art? Exploring the Movements That Define the Groundbreaking Genre," MY MODERN MET, 2022. https://mymodernmet.com/what-is-modern-art-definition/. [2] C. Ives, "Japonisme," New York: The Metropolitan Museum of Art, 2004. http://www.metmuseum.org/toah/hd/jpon/hd_jpon.htm. [3] 陳曼華, 藝術與文化政治：戰後台灣藝術的主體形構, 國立交通大學, 2016. [4] "Wikiart," https://www.wikiart.org/. [5] "American Realism," https://reurl.cc/RyRbLZ. [6] "Japonism," https://reurl.cc/RyRbLZ. [7] "Muralism," https://reurl.cc/ka9Lmd. [8] "Indian Space painting," https://reurl.cc/x6eEdz. [9] "Neo-baroque," https://reurl.cc/NyOZze. [10] Y. LeCun., Y. Bengio.,G. Hinton., "Deep learning," Nature, no. 521(7553), p. 436–444, 2015. [11] 簡秀枝, 台灣前輩油畫家市場之研究—以陳澄波、廖繼春、李梅樹、楊三郎、李石樵之油畫市場行情為例, 國立臺灣師範大學, 2007. [12] 倪再沁, "西方美術．台灣製造-台灣現代美術的批判," 雄獅美術, no. 242, pp. 132-133, 1991. [13] 劉聖秋, 70年代台灣鄉土美術之研究, 國立屏東師範學院, 2002. [14] 陸蓉之, "台灣地區當代藝術本土風格語彙的衍變," in 中華民國美術思潮研討會論文集, 台北市立美術館, 1992. [15] “臺北市立美術館,” https://www.tfam.museum/index.aspx?ddlLang=zh-tw. [16] “楊三郎美術館,” https://yangsanlang.com.tw/collection/. [17] “國立台灣美術館,” https://ntmofa-collections.ntmofa.gov.tw/Default.aspx. [18] E. Ahmed., M. Marian., B. Liu., K. Diana., E. Mohamed., The Shape of Art History in the Eyes of the Machine, arXiv, 2018. [19] A. Krizhevsky., I. Sutskever., G. E. Hinton., ImageNet Classification with Deep Convolutional Neural Networks, ACM, 2012. [20] S. Kumar., A. Tyagi., T. Sahu., P. Shukla., A. Mittal., Indian Art Form Recognition Using Convolutional Neural Networks, IEEE, 2018. [21] D. Kvak., Leveraging Computer Vision Application in Visual Arts: A Case Study on the Use of Residual Neural Network to Classify and Analyze Baroque Paintings, arXiv, 2022. [22] V. H. Phung., E. J. Rhee., A High-Accuracy Model Average Ensemble of Convolutional Neural Networks for Classification of Cloud Image Patches on Small Datasets, Hanbat National University, 2019. [23] “維基百科： Overfiting說明,” https://reurl.cc/x6o2mN. [24] K. Simonyan., A. Zisserman., Very Deep Convolutional Networks for Large-Scale Image Recognition, arXiv, 2014. [25] C. Szegedy., W. Liu., Y. Jia,P. Sermanet., S. Reed.,D. Anguelov.,A. Rabinovich., Going Deeper with Convolutions, IEEE, 2015. [26] K. He.,X. Zhang.,S. Ren.,J. Sun., Deep Residual Learning for Image Recognition, IEEE, 2016. [27] 李謦伊, “卷積神經網絡 CNN 經典模型 — LeNet、AlexNet、VGG、NiN with Pytorch code,” Medium, 2020. https://reurl.cc/q06LpE. [28] C. Shorte.,T. M. Khoshgoftaar., A Survey on Image Data Augmentation for Deep Learning. Journal of Big Data, SpringerOpen, 2019. [29] V. d. Maaten., G. Hinton., "Visualizing data using t-SNE," Journal of machine learning research, vol. 9, no. 11, 2008. [30] G. Hinton., E. Geoffrey., Roweis., T. Sam, "Stochastic neighbor embedding.," Advances in neural information processing systems., pp. 857-864, 2002. [31] 謝東山, “臺灣鄉土美術的質與量,” 臺灣美術期刊, 編號 107, p. 6~8, 2017. [32] 葉國新, 傳統與創新--由鄉土藝術出發探討繪畫創作的表現, 國立臺灣師範大學, 2005. [33] R. R. Selvaraju.,M. Cogswell.,A. Das,R. Vedantam.,D. Parikh.,D. Batra., Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization, IEEE, 2017. [34] “維基百科：浮世檜,” https://reurl.cc/Nyngm5. [35] “維基百科：墨西哥壁畫運動,” https://reurl.cc/8NzRZo. [36] “維基百科：巴洛克藝術,” https://reurl.cc/r6o2Gb. [37] "台灣鄉土美術運動," 維基百科, https://reurl.cc/gaNz6R. [38] 許綾讌, 舊園情懷 ─ 台灣鄉土古厝水墨畫之研究, 國立臺灣藝術大學, 2010. zh_TW
