Jul 2026· Communications in Science and Technology· Vol 11, pp. 138-149· 0 citations· 35 references
TL;DR
A hybrid framework for the classification of chili leaf disease that combines DenseNet201-based deep feature extraction with a Broad Learning System (BLS) classifier is investigated, indicating 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.
Abstract
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 are one of the major factors contributing to reduced agricultural productivity and economic losses for farmers. Manual disease identification generally requires expert knowledge and is often difficult to perform efficiently in field conditions. Therefore, this study aims to develop a rice leaf disease classification system by combining DenseNet201 as a feature extractor and a Voting Ensemble approach as the classifier. The dataset consisted of 1,470 rice leaf images categorized into five classes: Bacterial Leaf Blight, Brown Spot, Healthy Leaf, Leaf Blast, and Tungro. The dataset was divided using a stratified split strategy into 80% training data, 10% validation data, and 10% testing data. Image augmentation was applied only to the training set, increasing the number of training samples to 7,056 images. DenseNet201 was employed to extract image features into 1,920-dimensional feature vectors, which were subsequently classified using Logistic Regression, Support Vector Machine (SVM), Hard Voting, and Soft Voting. Experimental results showed that Logistic Regression achieved an accuracy of 95.24%, while SVM achieved 95.92%. Hard Voting obtained an accuracy of 95.24%, whereas Soft Voting achieved the best performance with an accuracy of 95.92%, precision of 95.75%, recall of 95.70%, F1-score of 95.71%, and ROC-AUC of 99.76%. Furthermore, the best-performing model was deployed in a Streamlit-based application for automatic rice leaf disease identification. The results demonstrate that the combination of DenseNet201 and Soft Voting provides an accurate and effective approach for rice leaf disease classification and has strong potential as an early disease detection tool in agriculture.
Nelly Khairani Daulay, Novi Lestari, R. Rusdiyanto· Jurnal Media Computer Scienc...· 0 citations
Agricultural productivity and food security are heavily impacted by plant diseases, and thus there is a high demand for accurate and automated plant disease detection that can be achieved by applying deep learning techniques. This research proposes a Multi-Model Ensemble Method Based on Deep Learning for multi-plant disease detection using ResNet50 to improve classification performance across multiple crop varieties. The proposed framework takes five important categories of plants into consideration including tomato, potato, grape, apple and maize, and 10 classes of healthy and diseased plants are generated from the PlantSeg dataset. The Anaconda platform was used along with Python to create a development environment that allows data preprocessing, augmentation, training and testing to be implemented efficiently. The proposed ensemble framework combines the feature extraction power of ResNet50 with several deep learning classifiers so as to obtain a good identification performance at different resolutions and environments. The proposed model performance is tested with the following metrics Accuracy, Precision, Inference Time, and Resolution quality and compared with MobileNetV2, YOLOv8 and the baseline CNN models. Experimental results show that the proposed ensemble ResNet50 framework achieves an accuracy of 98.7% and precision of 98.3%, which is about 6.4%, 4.8%, and 9.2% higher than that of MobileNetV2, YOLOv8, and CNN respectively. Moreover, the proposed method achieves high resolution disease localization capability with an inference time improvement of almost 18% compared with YOLOv8. The proposed system greatly improves the detection accuracy of the early stage and the calculation speed of the system, which is very suitable for smart agriculture applications and real-time monitoring of the health status of crops.
Objectives: To create a multi-class image classification system to automate the detection of potato crop diseases using deep learning algorithms to classify images of potato leaves. Method: This study involves an implementing and comparing of six deep learning models to classify potato leaves as diseased or infected with pests. The models included a custom CNN as the baseline and five transfer-learning models: VGG16, DenseNet121, MobileNetV2, Xception, and InceptionV3. The final selected model was InceptionV3 due to its ability to extract strong features and achieve superior overall classification performance among all evaluated models. To enhance model’s performance and improve generalization to unseen data, several techniques were implemented, including data augmentation, Batch Normalization, Dropout regularization, and selective fine-tuning of deeper layers. Findings: The proposed model achieved the highest test accuracy (94%) and macro-average F1-score (0.94) compared to other baseline models. The importance of fine-tuning is reflected in the high accuracy of the proposed model. An ablation study found that accuracy dropped to 84.67% without fine-tuning, which demonstrates how critical it is for this model’s domain adaptation. The Grad-CAM analysis showed that the model focuses on biologically relevant areas of the leaves with infection and does not concentrate on backgrounds; therefore, the results indicate the model’s potential for interpretability and deployment in real-world settings. Novelty: This study improves potato leaf disease detection using a fine-tuned InceptionV3 with data augmentation and dropout, while Grad-CAM visualizations enhance model interpretability, reliability, and practical utility for accurate agricultural disease diagnosis.
Keywords: PotatoLeaf Disease Detection, Deep Learning, Transfer Learning, InceptionV3, Image Classification, Grad-CAM, Sustainable Agriculture
Aradhy Tiwari, Amit Saxena, Chandrashekhar Chandrashekhar· Indian Journal of Science an...· 0 citations
The findings indicate that the combination of ResNet50 and Grad-CAM is effective for plant disease classification and provides better explainability for deep learning-based agricultural applications.
W. Zalmi, Rahmi Putri Kurnia, Dyah Listianing Tyas· Informatik : Jurnal Ilmu Kom...· 0 citations
Plants can suffer a number of diseases that impact agricultural productivity and food security, particularly in developing
farming communities. Although deep learning is capable of classification of diseases with outstanding results, its use is
limited due to the difficulty of obtaining large labeled databases and the high requirement of computation. To address
these challenges, this study proposes a new method for plant disease classification based on traditional image processing
and machine learning algorithms with lightweight and low computation requirements. This one uses several handcrafted
descriptors such as color histograms, Haralick texture features and Hu moments to retrieve the information relevant to a
disease from the segmented leaf images. Performance of top five classifiers, namely Random Forest, Support Vector
Machine, K-Nearest Neighbors, Logistic Regression and Naïve Bayes classifiers are evaluated from the dataset of
healthy plant leaves and diseased plant leaves images on 10-fold cross validation. Based on the results of the research
work, the best classification model was the Random Forest Classifier model with the accuracy value is 98.12%, 0.98
precision, 0.98 recall, and 0.98 F1 value. The proposed solution was also found to be uncomputation complex and low
memory consuming and can be made realtime inference. Therefore, this solution can be implemented onto agricultural
systems, considering edge computing and IoT. The results of the research also demonstrated that feature-based machine
learning approaches afford interpretable and reliable plant disease detection at a low computation cost, further
contributing to sustainable, and precision agriculture.
Pranay Meshram, Priyanka S. Meshram, Anshuman Kalia et al.· International Journal of Dru...· 0 citations
This research presents an automated detection method using the Single Shot Detector (SSD) framework, with ResNet-50 as the backbone and a Feature Pyramid Network (FPN) to manage multi-scale feature representations to strengthen plant disease monitoring systems.
Kusworo Adi, A. Setiadi, C. E. Widodo et al.· JOIV: International Journal...· 0 citations