Introduction Accurate and rapid diagnosis of plant leaf disease symptoms is critical for sustainable agricultural crop production, yet traditional methods often lack efficiency and robustness under field conditions. Methods Here, we propose a deep learning framework based on an improved Vision Transformer architecture that integrates a dynamic sparse attention mechanism, termed KBTNet, for targeted feature extraction in symptom-affected regions of leaf images. The model incorporates a learnable nonlinear enhancement module to capture subtle visual disease symptom variations such as lesions, discoloration patterns, and spot distributions, and a lightweight Transformer design to reduce computational cost. Results Evaluated on a multisource dataset containing soybean and tomato leaf images representing diverse disease symptom patterns, our approach achieved 93.19% classification accuracy, outperforming current state-of-the-art models. Additional evaluations on public plant disease datasets from multiple crops further demonstrate the model's ability to recognize disease symptom patterns across diverse crop species. Discussion The proposed framework achieves stable performance across diverse crop and disease symptom categories, maintains high efficiency under reduced parameter complexity, and exhibits strong potential for realtime field diagnostics on edge devices. This work provides a scalable and efficient tool for plant disease symptom detection and classification and supports the integration of visionbased intelligence into crop disease monitoring and management systems.
A unique, computationally efficient triple-feature block network capable of highly accurate plant disease classification across diverse species and complex imaging environments is proposed.
A. Elkholy, N. Elshennawy, Ahmed M. Gab Allah· Journal of King Saud Univers...· 0 citations
Potato leaf disease poses a significant threat to global food security, causing substantial crop losses that jeopardise agricultural productivity and farmers’ livelihoods worldwide. Existing automated detection frameworks suffer from several persistent limitations, including over-reliance on controlled benchmark datasets, narrow disease class coverage, exclusive use of spatial feature representations, absence of feature selection, and dependence on single-architecture end-to-end pipelines. To address these limitations, this paper proposes ComPo-Net, a novel lightweight ensemble framework that integrates three efficient CNN architectures—ResNet18, ShuffleNet, and MobileNetV2—for nine-class potato leaf disease detection and classification. Deep features are extracted from three intermediate layers of each network, with the Discrete Wavelet Transform applied for dimensionality reduction and cross-network fusion of the higher-dimensional layer features, capturing spectral–spatial information that purely spatial approaches cannot provide, while the remaining layer features are directly concatenated across networks. One-way Analysis of Variance (ANOVA) feature selection is subsequently applied to retain the most statistically significant features from the combined multi-scale, multi-network representation, and seven machine learning classifiers are systematically evaluated to identify the optimal classification strategy. The framework is assessed on a merged dataset of three publicly available benchmarks spanning both controlled and uncontrolled imaging environments, constituting a nine-class evaluation setting not previously addressed at this scale in the literature. ComPo-Net achieves an accuracy of 96.32%, an F1-score of 93.87%, an MCC of 0.9350, and AUC values exceeding 0.993 across all nine classes with Cubic SVM as the best-performing classifier. When compared against methods evaluated on the seven-class uncontrolled-environment dataset—the closest available task setting to ComPo-Net’s nine-class merged benchmark—ComPo-Net surpasses the best-performing comparable method by a margin of 6.45 percentage points, demonstrating the effectiveness of multi-scale ensemble feature extraction combined with spectral–spatial representation and principled feature selection for robust potato leaf disease detection under diverse real-world conditions.
Tomato, as a globally vital economic crop, relies heavily on accurate disease recognition to safeguard food security. However, tomato leaf disease identification constitutes a classic fine-grained visual classification task characterized by minimal inter-class variance, spatially sparse lesion features, and complex background interference. These challenges hinder conventional deep learning models from precisely localizing critical discriminative regions. In response to the aforementioned challenges, we introduce EA-HFA, an innovative framework based on deep neural networks that synergistically integrates an Entropy Attention mechanism alongside a Hierarchical Feature Alignment component. Specifically, the Entropy Attention module leverages information-theoretic entropy to quantify pixel-wise predictive uncertainty, adaptively selecting high-confidence pixels to automatically focus the network on sparse yet highly discriminative lesion features. Concurrently, the Hierarchical Feature Alignment module imposes KL-divergence constraints on the temperature-scaled probability distributions across adjacent network layers, enforcing cross-scale consistency in the localization of discriminative regions. Evaluations conducted on the PlantVillage and AI Challenger 2018 benchmarks reveal that EA-HFA achieves Top-1 accuracies of 99.29% and 97.82%, respectively, yielding performance comparable to established deep learning architectures while maintaining a reasonable computational footprint. Furthermore, qualitative analyses indicate that the model tends to attend to minute lesion-relevant areas, providing a certain level of interpretability for its decision-making process. Thus, EA-HFA holds practical potential as an alternative solution for automated plant disease monitoring in precision farming.
Coffee production plays a crucial role in the agricultural economy; however, its productivity is significantly affected by plant diseases that are difficult to detect at early stages. Accurate disease identification remains challenging due to subtle visual differences and high intra-class variability in leaf symptoms. To address this problem, this study proposes a hybrid deep learning framework that integrates Convolutional Neural Networks (CNN) and Data-efficient Image Transformers (DeiT) for automated coffee leaf disease classification. The proposed architecture leverages CNN to capture fine-grained local features, while DeiT models global contextual relationships through self-attention mechanisms, enabling a more comprehensive feature representation.
The model is trained and evaluated on a dataset of 6,048 labeled images across four classes: Healthy, Rust, Red Spider, and Leaf Miner. Experimental results demonstrate that the proposed CNN–DeiT model outperforms baseline CNN and Transformer-based approaches, achieving an accuracy of 93.1%, an F1-score of 92.3%, and a ROC-AUC of 95.6%. Robustness analysis shows that performance degradation remains limited (1.6%–3.4%) under various perturbation conditions, while out-of-distribution evaluation indicates strong generalization capability with only a minor accuracy decrease. These findings confirm that the hybrid CNN–Transformer architecture effectively enhances classification performance, robustness, and generalization. This study contributes to the advancement of deep learning methodologies in agricultural image analysis by providing a robust and scalable framework for plant disease classification, with potential applications in precision agriculture and data-driven crop management.
Early and accurate identification of plant diseases is essential for improving crop productivity and ensuring food security. Many existing deep learning-based plant disease classification methods rely solely on leaf images collected from a controlled environment, which limits their applicability in real-world agricultural conditions where symptoms may be visually unclear and influenced by environmental factors. To address these challenges, this study discusses AgriFusionNet, a context-aware multimodal deep learning framework that integrates leaf images, textual symptom descriptions, and environmental data for robust plant disease classification. The proposed architecture employs EfficientNet-B0 for visual feature extraction, BERT for semantic representation of symptom descriptions, and a lightweight multilayer perceptron for modeling environmental factors such as temperature, humidity, rainfall, and soil moisture. Features from all three modalities are fused into a unified representation to train the CNN model. The model is trained and tested upon the Context-Aware Multimodal Augmented PlantVillage dataset covering 38 plant diseases and healthy classes. Experimental results show that AgriFusionNet gives an overall accuracy of 98.94% on the dataset Context-Aware Multimodal Augmented PlantVillage, with competitive precision and recall and F1-score. The multimodal framework facilitates the co-learning of visual, semantic, and contextual environmental representations and the analyses of the confusion matrix and feature interactions give insights into cross-modal relationships. The proposed approach aims to explore context-aware multimodal representation learning for agricultural AI applications, with emphasis on integrating complementary visual, semantic, and contextual information.
V. C., Nischith N Shetty, M. Ramaiah et al.· Frontiers in Fungal Biology· 0 citations
—The existence of plant leaf diseases is a big problem for farmers all over the world because they make crops less healthy and less plentiful, which puts global food security at risk. The most common problems with diagnosing plant leaf diseases are a lack of experience, different ways of undertaking visual assessments, and image overlaps, all of which can lead to wrong diagnoses. This study introduces a hybrid deep learning architecture that integrates squeeze-and-excitation residual blocks, capsule networks, bidirectional long short-term memory, and attention mechanisms. The design utilizes convolutional operators for effective feature extraction, Squeeze-and-Excitation (SE) block for channel reweighting, capsule networks for spatial relationship capture Bidirectional Long Short-Term Memory (BiLSTM) for sequential dependencies, and attention mechanisms for emphasizing prominent features. Experiments were performed on 2 empirical datasets: the Corn Leaf Disease Dataset (CLDD) and the Rice Leaf Disease Dataset (RLDD). The data were divided into 60% for training, 20% for validation, and 20% for testing. The proposed method attained 99.88% training accuracy on CLDD and 100% on RLDD. During testing, the class-wise accuracies were 99.29% for blight and 100% for the other CLDD categories. In the case of RLDD, the accuracies attained were 78.95% for bacterial leaf blight, 85.53% for brown spot, 89.77% for healthy samples, 77.27% for leaf blast, 100% for leaf scald, and 97.73% for narrow brown spot. This work highlights practical potential for deployment in terms of mobile applications, enabling farmers to obtain rapid, reliable, and cost-effective field diagnoses, thereby improving agricultural productivity and sustainability.
Aekkarat Suksukont, Ekachai Naowanich· Journal of Advances in Infor...· 0 citations