EUVIMEDEUVIMED entdecken
Crossref

A Hybrid Deep Learning Framework Integrating Adaptive Patch Decomposition and Graph-based Disease Correlation Learning for Accurate Tomato Leaf Disease Classification

K. Kavitha, V. Selvi, K. Shunmuga Priya, R. Lakshmi

Indian Journal Of Agricultural Research · 21.07.2026 · (Of)

DOI: 10.18805/ijare.a-6561 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Indian Journal Of Agricultural Research, online erschienen am 21.07.2026. Verfasst von K. Kavitha und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Tomato (Solanum lycopersicum) leaf diseases, which reduce crop productivity, the quality of products and farm income, have a significant impact on agriculture worldwide. It is crucial to perform accurate, early identification of diseases to realize sustainable crop management and precise agriculture. Traditional manual diagnosis is costly, labor-intensive, inaccurate and subjective, which is unsuitable for large-scale agriculture. Though deep learning has been proven to be successful in classifying plant diseases, conventional methods have difficulties with background noises, inter-class similarity and insufficient learning of contextual features.

Methods

To overcome these issues, a hybrid deep learning framework based on adaptive patch decomposition (APD), graph-based disease correlation learning (GDCL) and hybrid decision fusion classification (HDFC) is proposed to efficiently and accurately classify tomato leaf diseases. The model was trained using 20,000 tomato leaf images with ten different diseases and healthy condition. First, image quality and training stability was enhanced by applying adaptive preprocessing techniques like image resizing, de-noising, contrast enhancement and normalization to the images. The ADPS module segmented lesions from irrelevant background information and generated the critical parts for disease-feature extraction. CNN-based local feature extraction and global context pooling were then performed to capture detailed lesion features and general leaf appearance, respectively. Graph-based disease correlation learning (GDCL) module was designed to establish inter-class disease correlation by using graph-based feature refine mechanism to differentiate between visually indistinguishable diseases. Lastly, hybrid decision fusion classification (HDFC) was proposed to enhance classification accuracy by combining prototype distance learning and graph similarity scores for disease classification.

Result

Experiment results showed that the proposed framework outperformed compared to other deep learning methods, including VGG16, ResNet50, DenseNet121, EfficientNet-B0 and InceptionV3, with the overall accuracy of 99.03%, precision of 98.98%, recall of 98.94%, F1-score of 98.78% and minimum loss of 0.053. The proposed system achieved robust stability and generalization ability tested with 5-fold cross validation. The model was proved to reduce inter-class confusion and obtain outstanding performance in learning the feature from lesions and it could be widely used in intelligent agricultural disease monitoring system and precise agriculture applications.

Zitieren

APA

Kavitha, K., Selvi, V., Priya, K. S., & Lakshmi, R. (2026). A Hybrid Deep Learning Framework Integrating Adaptive Patch Decomposition and Graph-based Disease Correlation Learning for Accurate Tomato Leaf Disease Classification. Indian Journal Of Agricultural Research. https://doi.org/10.18805/ijare.a-6561

Vancouver

Kavitha K, Selvi V, Priya KS, Lakshmi R. A Hybrid Deep Learning Framework Integrating Adaptive Patch Decomposition and Graph-based Disease Correlation Learning for Accurate Tomato Leaf Disease Classification. Indian Journal Of Agricultural Research. 2026. doi:10.18805/ijare.a-6561

BibTeX
@article{Kavitha2026,
  author = {Kavitha, K. and Selvi, V. and Priya, K. Shunmuga and Lakshmi, R.},
  title = {A Hybrid Deep Learning Framework Integrating Adaptive Patch Decomposition and Graph-based Disease Correlation Learning for Accurate Tomato Leaf Disease Classification},
  journal = {Indian Journal Of Agricultural Research},
  year = {2026},
  number = {Of},
  doi = {10.18805/ijare.a-6561}
}
Kostenlos starten →