| Titre : | Leveraging NLP for Plant Disease Detection in Smart Farms |
| Auteurs : | BITAM Sara, Auteur ; Adel Abdelli, Directeur de thèse |
| Type de document : | Mémoire magistere |
| Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2025 |
| Format : | 1 vol. (77 p.) / ill.couv.ill.encoul / 30cm |
| Langues: | Anglais |
| Langues originales: | Anglais |
| Résumé : |
Plant diseases compromise global crop production by 20–40%annually, threatening agricultural
sustainability and food security. In developing regions, farmers frequently lack timely access to phytopathological expertise and advanced diagnostic tools for effective disease management.This study presents an integrated AI-powered system for plant disease detection thatcombines visual analysis and treatment recommendation capabilities to support both individual farmers and agricultural enterprises. The system architecture implements a cascading multi-model approach: a MobileNetV2- based out-of-distribution (OOD) detection module validates input images as legitimate plant specimens with 98% accuracy. Valid images then proceed to an EfficientNet-B3 classifier that identifies the specific crop type (tomato, potato, or pepper) with 99% accuracy. Disease diagnosis is then performed by specialized models: DenseNet121 architectures for tomato and potato diseases, and a ResNet50 model for pepper diseases ensuring precise pathology identification across different plant families. The final component, a fine-tuned GPT-2 model, generates contextually appropriate treatment recommendations based on the diagnostic output, achieving strong linguistic performance metrics (ROUGE-1: 0.82, ROUGE-2: 0.73). The solution is deployed as a smartphone application. The system offers an analytical platform enabling agricultural companies to monitor disease prevalence patterns and assess treatment efficacy across geographical regions. This project aims to mitigate crop losses while promoting sustainable agricultural practices by delivering accessible AI-based disease detection tools and facilitating enhanced collaboration among farmers, researchers, and agricultural industry stakeholders. |
| Sommaire : |
Dedication i
Acknowledgements ii Abstract iii 1 General Introduction 1 1.1 Context and ProblemStatement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Motivation and Objectives . . . 2 1.3 Outline . . . . 3 2 Smart Agriculture and AI Applications 4 2.1 Introduction . . . . . 4 2.2 History of Farming . . 5 2.3 Traditional Agriculture .. 5 2.3.1 Characteristics of Traditional Farming . . . . . . . . . . . . . . . . . . . . . . 5 2.3.2 Traditional Farming Practices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3.3 Limitations of Traditional Agriculture . . . . . . . . . . . . . . . . . . . . . . . 6 2.3.4 Traditional Agriculture in theModern Context . . . . . . . . . . . . . . . . . . 7 2.4 Smart Agriculture . . .. 8 2.4.1 Traditional vs. Smart Farming . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.4.2 Technologies in Smart Farming . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.4.3 Advantages of Smart Agriculture . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.4.4 Smart Agriculture Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.5 Crop DiseaseManagement . . .12 2.6 AI Techniques in Agricultural Applications . . 12 2.6.1 Machine Learning . .13 2.6.2 Deep Learning . . 14 2.6.3 Classification vs. Clustering . . . . . 16 2.6.4 Natural Language Processing (NLP) . . . . . . . . . . . . . . . . . . . . . . . . 17 2.7 RelatedWorks . . . . . . . . . . . 18 2.8 Synthesis . . 21 2.9 Conclusion . . . 22 3 Design andMethodology 23 3.1 Introduction . . . . 23 3.2 Proposed Approach . . 23 3.3 System Functionalities . . 26 3.3.1 Use Case Analysis . . . .26 3.3.2 Activity Diagram Analysis . .28 3.4 Crop Disease Detection and Recommendation Process . . . . . . . . . . . . . . . . 29 3.4.1 Data Pipeline . . 31 3.4.2 Models Architecture . . . . 35 3.4.3 Performance Evaluation . . . . . . . . . . . . . 39 3.5 Conclusion .. . 41 4 Implementation and Results 42 4.1 Introduction . . . 42 4.2 Development Environment and Tools . 42 4.3 Models Development . .. 47 4.3.1 Out-of-Distribution (OOD) Detection Analysis and Results . . . . . . . . . . 47 4.3.2 Crop Classification Analysis and Results . . . . . . . . . . . . . . . . . . . . . 52 4.3.3 Tomato Disease Classification Analysis and Results . . . . . . . . . . . . . . . 58 4.3.4 Potato Disease Classification Analysis and Results . . . . . . . . . . . . . . . 63 4.3.5 Pepper Disease Classification Analysis and Results . . . . . . . . . . . . . . . 66 4.3.6 Dataset Size Impact . . . . . . . . . . . . . . . . 71 4.3.7 Recommendation Generation Analysis and Results . . . . . . . . . . . . . . . 72 4.4 Mobile Application Development and Results . . . . . . . . . . . . . . . . . . . . . . 77 4.5 Comparison Study . . 85 4.6 Conclusion . . . 87 5 Conclusion and Perspectives 88 References 90 |
| Type de document : | Mémoire master |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| MINF/927 | Mémoire master | bibliothèque sciences exactes | Consultable |
