A unique, computationally efficient triple-feature block network capable of highly accurate plant disease classification across diverse species and complex imaging environments is proposed.
The early detection of plant leaf diseases is essential for the enhancement of crop productivity and the promotion of sustainable agricultural practices. While deep learning models have been shown to be achieve remarkable success in the recognition of plant disease, conventional classifiers commonly rely on iterative gradient-based optimization, resulting in increased training complexity. This present study investigates a hybrid framework for the classification of chili leaf disease that combines DenseNet201-based deep feature extraction with a Broad Learning System (BLS) classifier. The DenseNet201 model is employed to generate discriminative feature representations, whereas the BLS approach employs a closed-form ridge regression solution for classification. The present study involved experiments conducted by means of the publicly available Chili Plant Leaf Disease Dataset containing 1,856 original images from six different disease categories. To prevent data leakage, the dataset was initially partitioned into training, validation, and test subsets at the original-image level, with data augmentation being applied exclusively to the training set, thereby increasing it to 9,093 images. The proposed DenseNet201+BLS framework achieved a test accuracy of 99.28% and a macro F1-score of 99.00%. Furthermore, the performance of the proposed model was compared with that of Softmax, Logistic Regression, Random Forest, Multilayer Perceptron, and Support Vector Machine (SVM) classifiers using identical DenseNet201 feature representations. Among the evaluated classifiers, SVM demonstrated the highest level of accuracy (99.64%), whereas BLS exhibited a favorable balance between predictive performance and computational efficiency, requiring less than one second for training while outperforming Softmax, Logistic Regression, and Random Forest. Grad-CAM visualizations further demonstrated that the extracted deep features focus on disease-relevant regions such as lesions, discoloration patterns, and abnormal leaf structures. The findings indicate that the integration of DenseNet201 feature extraction with a Broad Learning System offers a competitive and computationally efficient alternative for the automated classification of chili leaf disease. The proposed framework facilitates accurate disease recognition with substantially reduced training costs, making it a promising solution for resource-efficient agricultural monitoring and decision-support applications.
Rice leaf diseases significantly affect crop health and yield potential, creating a need for accurate and timely disease diagnosis. Many existing approaches still rely on single-stream feature extraction architectures, which may limit the ability to simultaneously capture global contextual information and fine-grained disease characteristics. Moreover, limited interpretability and decision-support capability hinder their practical deployment in real-world rice farming. To address these limitations, we employed a framework consists of two parallel feature extraction streams designed to capture different characteristics of disease patterns. The first stream uses a Swin Transformer to learn global contextual information and long-range spatial relationships across the leaf image. The second stream employs ConvNeXt to extract local texture features, including lesion details, spots, and color variations. By combining these complementary representations, the proposed framework effectively integrates global semantic information with local disease-specific features for improved classification performance. The extracted features from the two streams are fused through concatenation followed by an attention-based feature refinement module, enabling adaptive weighting of discriminative features. The refined representation is then used by a fully connected classifier for disease prediction. To enhance model interpretability, Grad-CAM visualization is incorporated to highlight disease-relevant regions and provide visual explanations for the model decisions. Furthermore, an LLM-based advisory module is integrated as a post-diagnosis decision-support component to provide contextualized disease management information and suggestions based on the predicted disease category. The generated suggestions are intended to support, rather than replace, expert agronomic recommendations and should be validated by agricultural professionals before practical application. The proposed framework was evaluated on two rice leaf disease datasets, achieving accuracies of 99.55% and 97.06% on Dataset-1 and Dataset-2, respectively, which are higher than those reported in previous studies. Additionally, five-fold cross-validation on Dataset-1 achieved an average accuracy of 98.91% ± 0.47, demonstrating the stability of the proposed approach. Cross-dataset evaluation using nine common disease classes across both datasets achieved 90.80% accuracy, indicating improved generalization across different data distributions. The proposed framework provides an accurate and explainable approach for rice leaf disease diagnosis in smart agriculture applications.
Existing classification systems of hybrid plant disease detection use CNN-based features extraction (e.g., VGG-16, ResNet-34, EfficientNet-B4) with SVM classification but are subject to major limitations (redundant features are extracted, features and classes are not well correlated, small datasets are overfitted, and low-quality images are poorly classified). Moreover, they are mostly limited to leaf-only analysis and do not have the validation on field data. To eliminate these problems, this paper introduces a developed hybrid model combining Vision Transformers, ConvNeXt, and EfficientNetV2 to extract features better. It uses the entropy-, ANOVA-, and mutual information-based feature selection to eliminate dimensionality and class relevance. Preprocessing of images enhances the capabilities to resist degraded input, and multi-part analysis of the plants can be used by real-life data that goes beyond leaves. The framework has lower computational cost, less overfitting, greater robustness to noise and complex backgrounds, greater F1-scores on imbalanced data and can be applied to mobile/IoT-based real-time disease diagnostic.
Automated detection and classification of orange diseases will greatly help to save money on fruit deterioration, improve fruit quality, and provide for long-term orange sustainability. Orange diseases usually appear as surface lesions or discoloration and can decrease the market price of the fruit and the likelihood of post-harvest decay. Since current inspection techniques depend heavily on the manual observations of trained inspectors, they are inefficient, biased, and cannot be applied to commercial-scale operations. Therefore, to address these issues, an automatic disease detection and classification framework using a Deep Convolutional Recurrent Neural Network (DCRNN) enabled by the optimization process of an Enhanced Sea Horse Optimization (ESHO) algorithm is developed. Pre-processing images with a Wiener filter to remove noise, CLAHE to amplify image contrast, and color-based segmentation to distinguish areas affected by disease via RGB thresholding is implemented. Next, deep feature extraction is achieved utilizing multiple pretrained convolutional models (i.e., ResNet50, VGG-16, and NasNet), which have different properties and are combined into one model using an entropy-based fusion technique. To achieve better performance by adjusting the hyperparameters of the DCRNN model, an Enhanced Sea Horse Optimization (ESHO) algorithm is utilized. Finally, classification of the disease is achieved using sophisticated machine learning algorithms such as SSAE, MHA-LSTM, and DCRNN. Additionally, Grad-CAM visualization was employed to enhance model interpretability by highlighting the disease-affected regions that influenced the classification decisions. Experiments were conducted using an open-source dataset for orange fruits, demonstrating that proposed ESHO-DCRNN framework produces better results than the traditional deep neural network approaches for detecting orange diseases, reaching 99.50% accuracy.
Plant diseases are crucial for improving crop yield and ensuring sustainable agricultural practices, particularly for staple crops such as groundnut and paddy leaf. However, existing methods often suffer from limited feature discrimination, inadequate attention to disease-affected regions, and reduced performance under real-world conditions. To address these limitations, this research introduces a novel deep learning (DL)-based GOPI-NET framework for precise groundnut and paddy leaf disease classification. Initially, the input leaf images are enhanced using Bilateral Filtering (BF) and Contrast Limited Adaptive Histogram Equalization (CLAHE) to reduce noise and improve contrast. Subsequently, HSV color space segmentation is employed to precisely isolate disease-affected regions. The proposed Dual Attention Network (DuAtNet) integrates channel and spatial attention mechanisms within a ConvNeXt backbone to capture discriminative disease-specific features. An efficient Fuzzy Extreme Learning Machine (FELM) classifier is then utilized for final categorization into Healthy, Leaf Spot, Bacterial Wilt, and Leaf Blight classes. The effectiveness of the GOPI-NET is evaluated using precision, recall, specificity, accuracy, and F1-score. The experimental results demonstrate that GOPI-NET achieves an overall accuracy of 98.32%. The GOPI-NET improves classification accuracy by 1.29%, 1.40%, and 2.23% compared to GLDICCNN, DNN-CSA, and LeafNet respectively.