Findings confirm that the hybrid CNN–Transformer architecture effectively enhances classification performance, robustness, and generalization in coffee leaf disease classification, with potential applications in precision agriculture and data-driven crop management.
Abstract
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.
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, 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
Introduction Maize is one of the most important food crops in the world, and foliar diseases can lead to significant yield losses if identification is not performed on time. Experts conducting manual inspections find it less effective and more subjective. Deep learning-based approaches utilizing Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) have been demonstrated as a viable approach to automate disease diagnosis. Thus, while CNNs fail to capture wider context due to their local feature focus and ViTs need larger datasets and tend to miss finer-grained details. To overcome these limitations, we present EDISP a hybrid framework that connects Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) for local feature extraction, as well as global contextual learning. Methods The EDISP framework brings together the strengths of CNNs and ViTs to overcome their individual weaknesses. It is trained on a dataset that includes both controlled-environment and real-field maize leaf images, which helps it handle different environmental conditions. The data undergoes thorough preprocessing, including normalization, augmentation, and stratified splitting into training, validation, and test sets to support generalization. The CNN focuses on detailed local disease features, while the ViT captures broader contextual information across the maize leaf surfaces. Results The proposed EDISP model significantly outperforms standalone CNN and ViT Models in multiple performance metrics, achieving an overall classification accuracy of 99.40%, precision of 99.43%, recall of 99.38%, and an F1-score of 99.40%. Experimental results demonstrate that EDISP excels in identifying maize leaf diseases, including Common Rust, Gray Leaf Spot, Northern Leaf Blight, and Healthy leaves, with minimal false positives and negatives. External validation with an independent dataset further highlights the model’s robustness and ability to generalize to real-world conditions. Discussion The EDISP hybrid architecture, integrating CNNs and ViTs, provides a stronger method for accurate, automated maize leaf disease detection. Its robust performance, consistent results on controlled and field datasets shows robustness in diverse environments. However, EDISP’s effectiveness may be limited by image quality, lighting, or disease types not seen in training. These results highlight the promise of hybrid deep learning in precision agriculture and offer a scalable solution for disease detection, supporting farmers without expert diagnostic resources.
Aziz Ullah, Shah Hussain, Qi Hou et al.· Frontiers in Plant Science· 0 citations
An extensive set of experiments was conducted to evaluate the performance of the proposed model for plant disease detection, and it is demonstrated that the model achieves highly reliable results, with an accuracy of 97.13%.
Hayat Meddeber, M. Meddeber· ITEGAM- Journal of Engineeri...· 0 citations
This study presents a novel CNN for multi-class classification of 38 diseases, demonstrating an effective balance between predictive performance and computational efficiency, positioning the model as a promising tool for real-world agricultural deployment.
A hybrid framework integrating a Convolutional Neural Network with a Large Language Model to perform image-based plant leaf disease classification accompanied by interpretable diagnostic explanations demonstrated that combining high-performance visual recognition with language-based reasoning enhanced both diagnostic reliability and interpretability in digital agriculture applications.
Frenky Riski Gilang Pratama, S. Surono, A. Thobirin· International Journal of Adv...· 0 citations
A deep hybrid Convolutional Neural Network –Transformer architecture is introduced by combining ConvNeXt Large (ConvNeXt-L) and Swin Transformer architecture by combining ConvNeXt Large (ConvNeXt-L) and Swin Transformer (as local feature extractor) and Swin Transformer (as global context encoder) to predict tomato leaf diseases.