To reduce losses from agriculture as well as enhance food security, we propose a three-stage deep ensemble for early citrus disease diagnosis from actual-field images of oranges (n = 2,240) as well as lemons (n = 208). To prevent leakage, augmentation is strictly enforced following splitting (70:30 stratified) following curation, normalisation, resizing by 224 × 224. Our contribution comes from combining state-of-the-art deep features by adding explicit texture priors. Namely, we add Local Binary Patterns (LBP) as well as Grey-Level Co-occurrence Matrix (GLCM) descriptors for micro-textures of lesions (e.g., stippling, scab rims around them, chlorosis encircled by veins) as well as second statistics (e.g., contrast, homogeneity, entropy) that CNNs/ViTs tend to discount by virtue of their small size coupled with variable-field data. These hand-crafted signals are z-score normalised as well as PCA-compressed for overfit protection as well as removal of collinearity then combined by deep embeddings. InceptionV3 (90% lemon) as well as DenseNet121 (93% orange) are the best of the five pretraining CNNs (ResNet50, DenseNet121, VGG16, InceptionV3, EfficientNetB0) that we test at Stage-1. The best CNNs are enlisted with a Vision Transformer (ViT) at Stage-2 for capture of long-range contextual capture improving upon Stage-1 by 98% (lemon) as well as 97% (orange). t-SNE confirms class separation while Stage-3 employs a multiclass SVM over the combined description that achieves 99% (lemon) while holding at 97% (orange) at another curation. The pipeline outperforms single-backbone variants, minimises variance while remaining lightweight enough for deployment, thus showing that LBP + GLCM texture priors compressed by PCA but combined by CNN/ViT features substantially enhance robustness plus generalisation for in-orchard citrus disease testing.

In numerous real-world applications, the class imbalance problem is prevalent. When training samples of one class immensely outnumber samples of the other classes, the traditional machine learning algorithms show bias towards the majority class (a class with more number of samples) lead to significant losses of model performance. Several techniques have been proposed to handle the problem of class imbalance, including data sampling and boosting. In this paper, we present a cluster-based oversampling with boosting algorithm (Cluster+Boost) for learning from imbalanced data. We evaluate the performance of the proposed approach with state-of-the-art methods based on ensemble learning like AdaBoost, RUSBoost and SMOTEBoost. We conducted experiments on 22 data sets with various imbalance ratios. The experimental results are promising and provide an alternative approach for improving the performance of the classifier when learned on highly imbalanced data sets.

Deep learning models are frequently implemented as black boxes, which restricts their application in sensitive sectors such as agriculture, despite their ability to achieve high classification accuracies. To resolve this issue, we introduce a comprehensive interpretability framework that implements Grad-CAM, Grad-CAM++, and LIME techniques on the most effective CNN + ViT + SVM ensemble model for the detection of citrus diseases. The proposed model obtained classification accuracies of 99% and 97% by augmenting the real-world Lemon and Orange datasets with 1600 and 4000 images, respectively. Interpretability was implemented on the DenseNet121 (Orange) and InceptionV3 (Lemon) backbones. The model focused on specific textures related to diseases, which was shown by the detailed image analysis from LIME, while Grad-CAM and Grad-CAM ++ highlighted clear areas of lesions with high detail. A perfect score of 1.00 over all assessed courses was achieved through quantified assessment using Intersection over Union (IoU) and Dice Coefficient, and confirmed accuracy in visual explanation. In addition, intra-class compactness and inter-class separability were confirmed using feature-space t-SNE mappings. This visual explanation framework enhances transparency and builds user trust in machine learning-based disease management decision making in practice through clearly interpretable visual explanations for both multi-class and binary classification tasks. This research is a major advance towards developing trustworthy and comprehensible AI (XAI) solutions for agricultural diagnosis, through providing visual explanations to both multi-class and binary classification tasks. The approach is scalable and model-independent and is capable of integration into real-time and mobile decision support tools.