Skip navigation

DSpace

機構典藏 DSpace 系統致力於保存各式數位資料(如:文字、圖片、PDF)並使其易於取用。

點此認識 DSpace
DSpace logo
English
中文
  • 瀏覽論文
    • 校院系所
    • 出版年
    • 作者
    • 標題
    • 關鍵字
  • 搜尋 TDR
  • 授權 Q&A
    • 我的頁面
    • 接受 E-mail 通知
    • 編輯個人資料
  1. NTU Theses and Dissertations Repository
  2. 電機資訊學院
  3. 資訊網路與多媒體研究所
請用此 Handle URI 來引用此文件: http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/7556
完整後設資料紀錄
DC 欄位值語言
dc.contributor.advisor張智星
dc.contributor.authorChih-Heng Hsiaoen
dc.contributor.author蕭至恆zh_TW
dc.date.accessioned2021-05-19T17:46:22Z-
dc.date.available2023-07-19
dc.date.available2021-05-19T17:46:22Z-
dc.date.copyright2018-07-19
dc.date.issued2018
dc.date.submitted2018-07-18
dc.identifier.citation[1] Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. ImageNet Large Scale Visual Recognition Challenge. Inter- national Journal of Computer Vision (IJCV), 115(3):211–252, 2015.
[2] Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. In Advances in neural information pro- cessing systems, pages 1097–1105, 2012.
[3] Sue Han Lee, Chee Seng Chan, Paul Wilkin, and Paolo Remagnino. Deep-plant: Plant identification with convolutional neural networks. CoRR, abs/1506.08425, 2015.
[4] Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbig- niew Wojna. Rethinking the inception architecture for computer vision. CoRR, abs/ 1512.00567, 2015.
[5] Christian Szegedy, Sergey Ioffe, and Vincent Vanhoucke. Inception-v4, inception- resnet and the impact of residual connections on learning. CoRR, abs/1602.07261, 2016.
[6] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. CoRR, abs/1512.03385, 2015.
[7] Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep net- work training by reducing internal covariate shift. CoRR, abs/1502.03167, 2015.
[8] François Chollet. Xception: Deep learning with depthwise separable convolutions. CoRR, abs/1610.02357, 2016.
[9] LaurentSifreandStéphaneMallat.Rigid-motion scattering for texture classification. CoRR, abs/1403.1687, 2014.
[10] François Chollet et al. Keras. https://keras.io, 2015.
[11] Martín Abadi, Ashish Agarwal, Paul Barham, Eugene Brevdo, Zhifeng Chen, Craig Citro, Greg S. Corrado, Andy Davis, Jeffrey Dean, Matthieu Devin, Sanjay Ghe- mawat, Ian Goodfellow, Andrew Harp, Geoffrey Irving, Michael Isard, Yangqing Jia, Rafal Jozefowicz, Lukasz Kaiser, Manjunath Kudlur, Josh Levenberg, Dande- lion Mané, Rajat Monga, Sherry Moore, Derek Murray, Chris Olah, Mike Schuster, Jonathon Shlens, Benoit Steiner, Ilya Sutskever, Kunal Talwar, Paul Tucker, Vincent Vanhoucke, Vijay Vasudevan, Fernanda Viégas, Oriol Vinyals, Pete Warden, Martin Wattenberg, Martin Wicke, Yuan Yu, and Xiaoqiang Zheng. TensorFlow: Large- scale machine learning on heterogeneous systems, 2015. Software available from tensorflow.org.
[12] Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. CoRR, abs/1409.1556, 2014.
[13] Pavel Golik, Patrick Doetsch, and Hermann Ney. Cross-entropy vs. squared error training: a theoretical and experimental comparison.
[14] Mei Wang and Weihong Deng. Deep visual domain adaptation: A survey. CoRR, abs/1802.03601, 2018.
[15] Jindong Wang et al. Everything about transfer learning and domain adapation. http://transferlearning.xyz.
[16] Barret Zoph, Vijay Vasudevan, Jonathon Shlens, and Quoc V. Le. Learning trans- ferable architectures for scalable image recognition. CoRR, abs/1707.07012, 2017.
dc.identifier.urihttp://tdr.lib.ntu.edu.tw/jspui/handle/123456789/7556-
dc.description.abstract本論文為植物影像的辨識與分類,利用深度神經網路 (deep neural network, DNN) 中的卷積神經網路去訓練出能對植物影像進行分類的模 型,論文裡將會使用起源網路 (InceptionV3)、起源殘差網路 (Inception- Resnet) 與極端起源網路 (Xception) 作為基礎模型進行訓練,訓練過程 將會使用資料增強 (data augmentation) 與遷移式學習 (transfer learning) 進行模型訓練,找出有最佳辨識率的模型進行下一階段器官獨立模型 的訓練。
第二階段為這篇論文提出的器官獨立模型的植物辨識方法,首先把 資料集依照器官標籤分成多個子資料集,利用上一階段訓練辨識正確 率最高的卷積神經網路模型,作為各自器官的分類模型,依照對應的 子資料集訓練出專精於各自植物器官的子分類模型,並再訓練一個器 官分類器,與多個子分類模型組合成器官獨立模型,嘗試使整體的辨 識率能夠再次上升,並對分類錯誤的資料進行分析。
zh_TW
dc.description.abstractThis paper focus on plant image recognition and classification. Briefly, we use extended deep neural network - convolution neural network(CNN) to train models for classifying plant images. For CNN model selection, Incep- tionV3, Inception-Resnet and Xception are very powerful condidates. These models can achieve state-of-the-art accuracy. First stage, we train these models via data augmentation and transfer learning. After experiment, model with the highest accuracy will be selected to the next stage.
In the next stage, trying to reduce error rate, we propose a new method called ”Organ Attribute Separated Model”. First of all, we divide the original dataset to organ separated datasets by organ labels. After plant subdataset generated, we will train multiple CNN models for every subdatasets and an organ classifier. Combining all these models to complete organ separated model. Last but not least, we will do error analysis by extracting features from CNN models.
en
dc.description.provenanceMade available in DSpace on 2021-05-19T17:46:22Z (GMT). No. of bitstreams: 1
ntu-107-R05944033-1.pdf: 14527155 bytes, checksum: 10831842c06730ab700cf1786e72a2bc (MD5)
Previous issue date: 2018
en
dc.description.tableofcontents口試委員會審定書 i
致謝 ii
摘要 iii
Abstract iv
目錄 v
圖表目錄 viii
表格目錄 xi
1 緒論 1
1.1 主題簡介.................................. 1
1.2 方法簡介.................................. 2
1.3 章節概述.................................. 2
2 研究方法 4
2.1 卷積神經網路 ............................... 4
2.1.1 卷積層............................... 4
2.1.2 池化層............................... 5
2.1.3 激活函數.............................. 6
2.1.4 全連接層.............................. 7
2.2 基於AlexNet的葉子辨識......................... 8
2.2.1 AlexNet模型架構......................... 9
2.3 起源神經網路....................... 10
2.3.1 1x1卷積核............................. 11
2.3.2 輔助分類器................. 12
2.3.3 卷積核空間分解.......................... 12
2.3.4 標籤平滑化.................. 13
2.4 起源殘差神經網路............................. 14
2.4.1 區段正規化 ............................ 14
2.4.2 起源殘差網路架構 ........................ 15
2.5 極端起源神經網路............................. 17
2.5.1 極端起源網路架構 ........................ 17
2.5.2 分離式卷積 ............................ 18
2.5.3 極端神經網路的其他嘗試 .................... 20
2.6 損失函數.......................... 21
2.6.1 均方誤差.............................. 21
2.6.2 交叉熵............................... 23
3 實驗設置與結果 24
3.1 資料集 ................................... 24
3.1.1 Top-500資料集 .......................... 25
3.1.2 Top-500器官資料集........................ 26
3.2 實驗環境.................................. 28
3.3 影像前處理................................. 29
3.3.1 重縮放............................... 29
3.3.2 資料增強.............................. 30
3.4 遷移式學習與模型選取.......................... 33
3.4.1 監督式影像分類的遷移式學習.................. 33
3.4.2 卷積神經網路模型選取...................... 35
3.4.3 遷移式學習特徵圖群分析 .................... 41
3.5 器官獨立模型 ............................... 43
3.5.1 器官子模型訓練.......................... 44
3.5.2 器官分類器訓練.......................... 46
3.5.3 整合器官獨立模型 ........................ 48
3.6 錯誤分析.................................. 50
4 結論與未來展望 53
4.1 結論..................................... 53
4.1.1 辨識率............................... 53
4.1.2 實際應用.............................. 54
4.2 未來展望.................................. 54
參考文獻 56
dc.language.isozh-TW
dc.title利用遷移式學習與植物器官獨立模型進行植物影像辨識zh_TW
dc.titlePlant Image Recognition by Transfer Learning and Plant Organ Separated Modelen
dc.typeThesis
dc.date.schoolyear106-2
dc.description.degree碩士
dc.contributor.oralexamcommittee徐宏民,王鈺強
dc.subject.keyword植物影像辨識,卷積神經網路,遷移式學習,特徵向量抽取,起源神經網路,分散式模型,zh_TW
dc.subject.keywordPlant image recognition,Convolution neural network,Transfer learning,Feature extraction,Inceptin network,Separated model,en
dc.relation.page57
dc.identifier.doi10.6342/NTU201801651
dc.rights.note同意授權(全球公開)
dc.date.accepted2018-07-18
dc.contributor.author-college電機資訊學院zh_TW
dc.contributor.author-dept資訊網路與多媒體研究所zh_TW
dc.date.embargo-lift2023-07-19-
顯示於系所單位:資訊網路與多媒體研究所

文件中的檔案:
檔案 大小格式 
ntu-107-1.pdf14.19 MBAdobe PDF檢視/開啟
顯示文件簡單紀錄


系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。

社群連結
聯絡資訊
10617臺北市大安區羅斯福路四段1號
No.1 Sec.4, Roosevelt Rd., Taipei, Taiwan, R.O.C. 106
Tel: (02)33662353
Email: ntuetds@ntu.edu.tw
意見箱
相關連結
館藏目錄
國內圖書館整合查詢 MetaCat
臺大學術典藏 NTU Scholars
臺大圖書館數位典藏館
本站聲明
© NTU Library All Rights Reserved