EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

JackVisualNet: a fine-tuned hybrid deep learning model for jackfruit disease classification with explainable AI

Amir Sohel, Md. Hasan Imam Bijoy, Sarbajit Paul Bappy, Rittik Chandra Das Turjy, Manal Othman, Md Abdus Samad

PeerJ Computer Science · 07.07.2026 · 12 · e3977

DOI: 10.7717/peerj-cs.3977 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PeerJ Computer Science, online erschienen am 07.07.2026. Verfasst von Amir Sohel und 5 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Jackfruit, a vital agricultural crop in Bangladesh, is a key player in ensuring food security and sustaining rural communities’ livelihoods. The escalating challenges posed by plant diseases and the shortcomings of traditional manual disease detection methods underscore the pressing need for an automated, scalable solution. This study introduces JackVisualNet, a cutting-edge hybrid deep learning model tailored for automated detection and classification of jackfruit diseases. This innovative system, which integrates advanced technologies, including pre-trained deep learning architectures, hybrid model design, and intelligent classification algorithms, has the potential to significantly impact food security in Bangladesh. This study leverages a dataset of 2,195 high-resolution images collected from jackfruit fields in Barishal and Khulna, Bangladesh, encompassing diverse disease conditions. Data preprocessing included resizing images to standardized dimensions (224 × 224 × 3) and applying augmentation techniques to improve model robustness. JackVisualNet combines the strengths of VGG19 and DenseNet201, further refined through extensive hyperparameter tuning and a robust ablation study. This hybrid architecture optimally enhances feature extraction and classification, addressing variability in disease prediction. Benchmarking against ResNet50V2, DenseNet201, and VGG19 demonstrates that JackVisualNet achieves superior performance, with accuracies of 99.72%, 99.73%, and 99.72%, precision of 99.73%, 99.73%, and 99.73%, recall of 99.72%, 99.72%, and 99.72%, and an F1-score of 99.73%, 99.73%, and 99.73%, respectively. Beyond classification, the system incorporates predictive analytics for disease-spread trends and a decision-based algorithm that provides actionable insights for farmers. The results highlight JackVisualNet’s capability to revolutionize jackfruit disease management, offering an accessible, scalable, and reliable solution for sustainable agricultural practices. JackVisualNet’s reliability provides a secure and confident solution for farmers, empowering them with actionable insights and promoting efficient disease management, thereby contributing to the resilience of Bangladesh’s agricultural sector.

Förderung

  • Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia

Zitieren

APA

Sohel, A., Bijoy, M. H. I., Bappy, S. P., Turjy, R. C. D., Othman, M., & Samad, M. A. (2026). JackVisualNet: a fine-tuned hybrid deep learning model for jackfruit disease classification with explainable AI. PeerJ Computer Science, 12, e3977. https://doi.org/10.7717/peerj-cs.3977

Vancouver

Sohel A, Bijoy MHI, Bappy SP, Turjy RCD, Othman M, Samad MA. JackVisualNet: a fine-tuned hybrid deep learning model for jackfruit disease classification with explainable AI. PeerJ Computer Science. 2026;12:e3977. doi:10.7717/peerj-cs.3977

BibTeX
@article{Sohel2026,
  author = {Sohel, Amir and Bijoy, Md. Hasan Imam and Bappy, Sarbajit Paul and Turjy, Rittik Chandra Das and Othman, Manal and Samad, Md Abdus},
  title = {JackVisualNet: a fine-tuned hybrid deep learning model for jackfruit disease classification with explainable AI},
  journal = {PeerJ Computer Science},
  year = {2026},
  volume = {12},
  pages = {e3977},
  doi = {10.7717/peerj-cs.3977}
}
Kostenlos starten →