| Titre : | Incorporating Deep Learning and Optimization Techniques with Data Augmentation for Improved Image Analysis and Classification |
| Auteurs : | Nouara Boudouh, Auteur ; Bilal Mokhtari, Directeur de thèse |
| Type de document : | Thése doctorat |
| Année de publication : | 2025 |
| Format : | 1 vol. (124 p.) |
| Langues: | Anglais |
| Langues originales: | Anglais |
| Mots-clés: | Deep Learning ; Optimization ; Image Analysis |
| Résumé : |
Deep learning methods often face challenges due to unbalanced or non-representative data, and in many cases, data scarcity limits model effectiveness. We advocate that improving data quality can lead to significant performance enhancements. This thesis presents new methods for data augmentation. Our first method involves randomly create filters to remove certain rows and columns from the original image to generate smaller, more informative images. This method was applied to the Cats vs. Dogs dataset to train the Basic CNN and ResNet50 models, showing improved results compared to the original dataset. However, random filter generation can sometimes produce images that are too similar to the originals, reducing diversity. To address this, we developed a secondary technique incorporating a random optimization algorithm to select optimal generated images based on entropy, yielding promising results when applied to the VGG16 model. Nevertheless, image selection remains dependent on filter quality, potentially limiting diversity. Therefore, our third method employs a genetic algorithm to enhance filter generation and ensure greater diversity. Additionally, we improved the architectures of the VGG16 and VGG19 models. When applied to the Cats vs. Dogs and Chest X-ray datasets and used to train a set of seven models (VGG16, VGG19, their enhanced versions, EfficientNet-B0, Inception-V3, and Vision Transformer), we observed promising improvements in model performance compared to the second method. Since optimization techniques require considerable time and resources, we proposed an alternative method to enhance model performance without increasing data size. This approach leverages the unique capabilities of each model to extract features by merging their outputs into a unified representation used to train a single classifier. The integrated models using VGG16, VGG19, EfficientNet-B0, and Inception-V3 showed clear performance superiority compared to each model’s individual performance. |
| Sommaire : |
Acknowledgement I Dedication II Abstract III Résumé IV List of Figures IX List of Tables XII List of Algorithms XIII List of Abbreviations XV List of Publications XVI General Introduction 10 1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2 Research Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1 Fundamental Concepts and Related Work 16 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.2 Image classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 1.3 Data Augmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 1.3.1 Classic approaches . . . . . . . . . . . . . . . . . . . . . . . . . . 20 1.3.2 Deep learning-based DA approaches . . . . . . . . . . . . . . . . 26 1.4 Challenges and Difficulties . . . . . . . . . . . . . . . . . . . . . . . . . . 33 1.4.1 Limitations of traditional approaches . . . . . . . . . . . . . . . . 34 1.4.2 Limitations of deep learning-based approaches . . . . . . . . . . . 35 1.5 Optimization methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 1.5.1 Exact methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 1.5.2 Heuristic and metaheuristic methods . . . . . . . . . . . . . . . . 36 1.5.3 Convex optimization . . . . . . . . . . . . . . . . . . . . . . . . . 39 1.5.4 image optimization-based classification . . . . . . . . . . . . . . . 39 1.6 Experimental Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 1.6.1 Cats vs. Dogs dataset . . . . . . . . . . . . . . . . . . . . . . . . 41 1.6.2 Chest X-ray dataset . . . . . . . . . . . . . . . . . . . . . . . . . 41 1.7 Experimental Deep Learning Architectures . . . . . . . . . . . . . . . . . 42 1.7.1 Basic CNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 1.7.2 ResNet50 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 1.7.3 VGG16 and VGG19 . . . . . . . . . . . . . . . . . . . . . . . . . 44 1.7.4 The enhanced VGG16 and VGG19 . . . . . . . . . . . . . . . . . 44 1.7.5 Inception-V3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 1.7.6 The Vision Transformer (ViT) . . . . . . . . . . . . . . . . . . . . 46 1.7.7 EfficientNet-B0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 1.8 Evaluation Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 1.8.1 Confusion matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 1.8.2 Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 1.8.3 Error rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 1.8.4 Recall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 1.8.5 F1 score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 1.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 2 Random Pixel Selection through Image Cropping for Data Augmentation and Classification 51 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 2.2 Proposed Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 2.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 3 Random Optimization and Entropy-Based DA for Image Classification and Analysis ”ROEDA” 64 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.2 Random Optimization Method . . . . . . . . . . . . . . . . . . . . . . . 66 3.3 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 3.4 Proposed Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 3.5 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4 Enhancing Deep Learning Image Classification Using Data Augmentation and Genetic Algorithm-based Optimizations 76 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 4.2 Proposed Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.2.1 Generate filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 4.2.2 Application of GA . . . . . . . . . . . . . . . . . . . . . . . . . . 81 4.2.3 Image generation . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 4.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5 Enhancing Image Classification with Ensemble Deep Learning through Deep Feature Concatenation 108 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 5.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 5.3 Proposed Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 5.3.1 Feature extraction using CNN . . . . . . . . . . . . . . . . . . . . 112 5.3.2 Concatenation of feature vectors . . . . . . . . . . . . . . . . . . 114 5.3.3 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 5.4 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 5.5 Challenges of Concatenation Method . . . . . . . . . . . . . . . . . . . . 118 5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 Conclusion and Future Works 121 1 Summary and key findings . . . . . . . . . . . . . . . . . . . . . . . . . . 122 2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 |
| En ligne : | http://thesis.univ-biskra.dz/id/eprint/6825 |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| TINF/199 | Théses de doctorat | bibliothèque sciences exactes | Consultable |
