使用DenseSwin Transformer 深度學習模型對早期和晚期玉米植株病理的分類:一個尚比亞農民的應用研究

簡易檢索 / 詳目顯示

回結果列表

研究生：	祈祖 Chiuzu Chilumbu
論文名稱：	使用DenseSwin Transformer 深度學習模型對早期和晚期玉米植株病理的分類:一個尚比亞農民的應用研究 Utilizing a DenseSwin Transformer Deep Learning Model to Classify Maize Plant Pathology in Early and Late Stages: A Study on Zambian Farmers’ Application
指導教授：	孫宏民 Sun, Hung-Min
口試委員:	許富皓 Hsu, Fu-Hau 黃育綸 Huang, Yu-Lun
學位類別：	碩士 Master
系所名稱：	電機資訊學院 - 資訊系統與應用研究所 Institute of Information Systems and Applications
論文出版年：	2023
畢業學年度：	111
語文別：	中文
論文頁數：	60
中文關鍵詞：	玉米病理學、Transformer 架構、玉米病害分類、DenseSwin 、多頭自注意力機制
外文關鍵詞：	Maize Pathology, Transformer Architecture, Maize disease classification, DenseSwin, Multi-head Self-Attention Mechanism
相關次數：	點閱：2 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

玉米是許多撒哈拉地區以南國家的主食作物，包括尚比亞，它容易受到當地多種疾病的威脅而造成重大影響。為了解決這個挑戰並提高疾病檢測效率，深度學習方法被使用來準確分類和識別植物疾病。最近，在尚比亞的許多地區，手動檢查玉米種植區以進行疾病檢測已成為常態。但這個過程不僅耗時，而且對於大規模農業運作來說也不實際效益。因此，在現代農業中，精確且自動化類別分類模型變得非常重要。在本研究中，我們提出了一種稱作DenseSwin的新型深度學習模型，專門應用在玉米疾病在早期和晚期的分類。DenseSwin結合了密集連接的卷積層和平移窗口的多重自注意機制的優勢。這種機制融合了模型能夠有效識別玉米植物圖像中的複雜模態和特徵，從而提高疾病分類性能。透過嚴格的實驗與評估結果，DenseSwin表現了97.18%的效能準確度。這些結果突顯了該模型準確檢測和分類玉米疾病的出色能力，可在農業領域，尤其是在尚比亞地區的農業實際應用提供了卓越的潛力。

Maize, which is the primary crop in many sub-Saharan countries, including Zambia, is susceptible to a wide range of diseases that have a significant impact on food production. To tackle this challenge and improve disease detection efficiency, deep learning methods have been employed to accurately classify and identify plant diseases. In recent times, manual inspection of maize fields for disease detection has been the standard practice in many parts of Zambia. However, this approach is not only time-consuming but also impractical for large-scale agricultural operations. Hence, the development of precise and automated classification models has become crucial in modern agriculture. In this study, we propose a novel deep-learning model called DenseSwin, specifically designed for maize disease classification in both the early visible stage and late indisputable stage of the disease. DenseSwin combines the strengths of densely connected convolution blocks with a shifted windows-based multi-head self-attention mechanism. This unique fusion of techniques enables the model to effectively capture intricate patterns and features in maize plant images, thereby enhancing disease classification performance. Through extensive experimentation and evaluation, DenseSwin achieves an impressive accuracy of 97.18%. These results highlight the model’s remarkable ability to accurately detect and classify maize diseases, offering promising potential for real-world applications in agricultural settings, particularly in Zambia.

Table of Contents
Abstract (Chinese) I
Abstract II
Acknowledgements III
Contents IV
List of Figures VII
List of Tables IX
 Introduction 1  2  Background and Motivation 5
1  Climate change impact on agriculture in Zambia. . . . . . . . . . . 5
2  Maize plant pathology ......................... 6
2.1 Pest ............................... 7
2.2 Viral............................... 8
2.3 Bacterial............................. 10
3  Need for automation .......................... 13  3  Related Works 15  3.1 Adopting machinery for agriculture .................. 15
2  Automation for maize diseases classification . . . . . . . . . . . . . 16
3  Deep Learning for maize classification. . . . . . . . . . . . . . . . . 17
3.1  CNN for maize disease classification. . . . . . . . . . . . . . 18  3.3.2  Vision Transformers for plant disease classification . . . . . . 20  3.3.3  Hybrid models for plant disease classification . . . . . . . . . 22  4 Methodology 23
1  Data Collection ............................. 24
1.1  KaraAgro AI Maize Dataset ................. 25  4.1.2  Makerere University Maize Image Dataset . . . . . . . . . . 25  4.1.3  Namibia University of Science and Technology maize Dataset 26  4.1.4  Tanzania Agricultural Research Institute Dataset . . . . . . 27  4.1.5  Maize Leaf Disease Dataset .................. 27  4.2  Data Preparation ............................ 29  4.3  DenseSwin Framework Overview.................... 29
3.1 Convolution Structure ..................... 30
3.2 Transformer Structure ..................... 32  4.4  DenseSwin Model Architecture .................... 34
4.1 Pooling Layer.......................... 35
4.2 Individual stage......................... 36  5 Experiment Results 41
1  Experiment Setup............................ 41  5.2  Results.................................. 42
2.1 Performance Analysis...................... 43
2.2 Visual Analysis ......................... 48  5.3  Discussion................................ 50  6 Conclusion 52
1 Challenges and Limitations ...................... 53
2 Future Work............................... 53
Bibliography 55

                                

[1] Darlington Akogo, Issah Samori, Christabel Acquaye, Michael Addo, Emmanuel Amoako, Frank Ezroa-Cudjoe, Jerry Buaba, Tatu Nyaku Seloame, Agyeiwaa Mavis, Hodasi Bright, Gyasi-Darko Kezia, Yao Owusu Ababio, and Darkwa Asare Clinton. The KaraAgro AI Maize Dataset. 2022.  
[2] Claire Babirye, Joyce Nakatumba-Nabende, Gloria Namanya, Chodrine Mutebi, Moses Ebellu, Joab Murungi, Saolo Tobius, Jonah Ssemwogerere, Annet Nakayima, Deborah Nabagereka, Judith Asasira, and Ruth Kanyesigye. Makerere University Maize Image Dataset. Harvard Dataverse, 2022.  
[3] Jayme Garcia Arnal Barbedo. A review on the main challenges in automatic plant disease identification based on visible range images. Biosystems Engineering, 144:52–60, 2016.  
[4] Jayme Garcia Arnal Barbedo. Impact of dataset size and variety on the effectiveness of deep learning and transfer learning for plant disease classification. Computers and Electronics in Agriculture, 153:46–53, 2018.  
[5] Akshaya Kumar Biswal, Amos Emitati Alakonya, Khondokar Abdul Mottaleb, Sarah J Hearne, Kai Sonder, Terence Luke Molnar, Alan M Jones, Kevin Vail Pixley, and Boddupalli Maruthi Prasanna. Maize lethal necrosis disease: review of molecular and genetic resistance mechanisms, socio-economic impacts, and mitigation strategies in sub-saharan africa. BMC Plant Biol., 22(1):542, November 2022.

[6] Sibanda Blessing, A.M Gamundani, G.E Iyawa, Lonnie S. Matsa, Alphons K. Koruhama, Josephine T. Pasipanodya, Belly Kasaona, Albertina Kadhikwa, and Helena N Amadhilla. Namibia University of Science and Technology Maize Dataset. 2022.  
[7] Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale, 2021.  
[8] Shundong Fang, Yanfeng Wang, Guoxiong Zhou, Aibin Chen, Weiwei Cai, Qifan Wang, Yahui Hu, and Liujun Li. Multi-channel feature fusion networks with hard coordinate attention mechanism for maize disease identification under complex backgrounds. Computers and Electronics in Agriculture, 203:107486, 2022.  
[9] Konstantinos P. Ferentinos. Deep learning models for plant disease detection and diagnosis. Computers and Electronics in Agriculture, 145:311–318, 2018.  
[10] Sohum Gupta, Josh Vishnoi, and A.S. Rao. Disease detection in maize plant using deep convolutional neural network. In 2022 7th International Conference on Communication and Electronics Systems (ICCES), pages 1330–1335, 2022.  
[11] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition, 2015.  
[12] Andrew Howard, Mark Sandler, Grace Chu, Liang-Chieh Chen, Bo Chen, Mingxing Tan, Weijun Wang, Yukun Zhu, Ruoming Pang, Vijay Vasudevan, Quoc V. Le, and Hartwig Adam. Searching for mobilenetv3, 2019.  
[13] Gao Huang, Zhuang Liu, Laurens van der Maaten, and Kilian Q. Weinberger. Densely connected convolutional networks, 2018.  
[14] David. P. Hughes and Marcel Salathe. An open access repository of images on plant health to enable the development of mobile disease diagnostics, 2016.  
[15] Sahil Jasrotia, Jyotsna Yadav, Navin Rajpal, Mukta Arora, and Juhi Chaudhary. Convolutional neural network based maize plant disease identification. Procedia Computer Science, 218:1712–1721, 2023. International Conference on Machine Learning and Data Engineering.  
[16] Chapwa Kasoma, Hussein Shimelis, Mark D. Laing, Admire Shayanowako, and Isack Mathew. Outbreaks of the fall armyworm (spodoptera frugiperda), and maize production constraints in zambia with special emphasis on coping strategies. Sustainability, 13(19), 2021.  
[17] Xiaopeng Li, Xiaoyu Chen, Jialin Yang, and Shuqin Li. Transformer helps identify kiwifruit diseases in complex natural environments. Computers and Electronics in Agriculture, 200:107258, 2022.  
[18] Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows, 2021.  
[19] Jerry Ma and Denis Yarats. On the adequacy of untuned warmup for adaptive optimization, 2021.  
[20] R. Manavalan. Automatic identification of diseases in grains crops through computational approaches: A review. Computers and Electronics in Agriculture, 178:105802, 2020.  
[21] Neema Mduma, Hudson Laizer, Loyani Loyani, Mbwana Macheli, Zablon Msengi, Alice Karama, Irine Msaki, Sophia Sanga, and Kennedy Jomanga. The Nelson Mandela African Institution of Science and Technology Maize dataset. 2022.  
[22] Blessing Mhlanga, Mulundu Mwila, and Christian Thierfelder. Improved nutrition and resilience will make conservation agriculture more attractive for zambian smallholder farmers. Renewable Agriculture and Food Systems, 36(5):443–456, 2021.  
[23] I. Mukumbuta, L.M. Chabala, S. Sichinga, and R.M. Lark. Accessing and assessing legacy soil information, an example from two provinces of zambia. Geoderma, 420:115874, 2022.  
[24] Wilson Nguru and Caroline Mwongera. Predicting the future climate-related prevalence and distribution of crop pests and diseases affecting major food crops in zambia. PLOS Climate, 2:1–25, 01 2023.  
[25] Diane Niyomwungere, Waweru Mwangi, and Richard Rimiru. Multi-task neural networks convolutional learning model for maize disease identification. In 2022 IST-Africa Conference (IST-Africa), pages 1–9, 2022.  
[26] Serosh Karim Noon, Muhammad Amjad, Muhammad Ali Qureshi, and Abdul Mannan. Use of deep learning techniques for identification of plant leaf stresses: A review. Sustainable Computing: Informatics and Systems, 28:100443, 2020.  
[27] Godfrey Omulo, Regina Birner, Karlheinz Ko ̈ller, Simunji Simunji, and Thomas Daum. Comparison of mechanized conservation agriculture and conventional tillage in zambia: A short-term agronomic and economic analysis. Soil and Tillage Research, 221:105414, 2022.  
[28] The CIMMYT Maize Program. Maize diseases: A guide for field identification. 2004.  
[29] Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-CAM: Visual explanations from deep networks via gradient-based localization. International Journal of Computer Vision, 128(2):336–359, oct 2019.  
[30] Vivek Sharma, Ashish Kumar Tripathi, and Himanshu Mittal. Technological revolutions in smart farming: Current trends, challenges future directions. Computers and Electronics in Agriculture, 201:107217, 2022.  
[31] Joa ̃o Vasco Silva, Fr ́ed ́eric Baudron, Hambulo Ngoma, Isaiah Nyagumbo, Esau Simutowe, Kelvin Kalala, Mukwemba Habeenzu, Mtendere Mphatso, and Christian Thierfelder. Narrowing maize yield gaps across smallholder farming systems in zambia: what interventions, where, and for whom? Agron. Sustain. Dev., 43(2), April 2023.  
[32] Davinder Singh, Naman Jain, Pranjali Jain, Pratik Kayal, Sudhakar Kumawat, and Nipun Batra. PlantDoc. In Proceedings of the 7th ACM IKDD CoDS and 25th COMAD. ACM, jan 2020.  
[33] Mingxing Tan and Quoc V. Le. Efficientnet: Rethinking model scaling for convolutional neural networks, 2020.  
[34] Poornima Singh Thakur, Pritee Khanna, Tanuja Sheorey, and Aparajita Ojha. Trends in vision-based machine learning techniques for plant disease identification: A systematic review. Expert Systems with Applications, 208:118117, 2022.  
[35] Christian Thierfelder, Fr ́ed ́eric Baudron, Peter Setimela, Isaiah Nyagumbo, Walter Mupangwa, Blessing Mhlanga, Nicole Lee, and Bruno G ́erard. Complementary practices supporting conservation agriculture in southern africa. a review. Agron. Sustain. Dev., 38(2), April 2018.

[36] Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Herv ́e J ́egou. Training data-efficient image transformers distillation through attention, 2021.  
[37] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need, 2017.  
[38] Wei Wenxuan, Wang Qianshu, Hao Chaofan, Sun Xizhe, Bao Ruiming, and Teoh Teik Toe. Leaf disease image classification method based on improved convolutional neural network. In 2022 IEEE International Conference on Industry 4.0, Artificial Intelligence, and Communications Technology (IAICT), pages 210–216, 2022.  
[39] Yanlei Xu, Bin Zhao, Yuting Zhai, Qingyuan Chen, and Yang Zhou. Maize diseases identification method based on multi-scale convolutional global pooling neural network. IEEE Access, 9:27959–27970, 2021.  
[40] Zambia and FAO. Partnering for improved livelihoods and climate change mitigation. Project implemented by FAO in close collaboration with the Ministry of Agriculture (MoA). Funded by the European Union (EU), pages 1–2, 2017.

簡易檢索 / 詳目顯示

相關論文