EUVIMEDEUVIMED entdecken
Crossref

Deep hierarchical spatial–sequential framework for robust multiclass leaf disease analytics

Monika, Ankit Kumar

IET Conference Proceedings · 01.10.2026 · 2026 (15) · 674-678

DOI: 10.1049/icp.2026.3973 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IET Conference Proceedings, erschienen am 01.10.2026. Verfasst von Monika und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Agricultural productivity suffers greatly because crop diseases are difficult to detect with proper accuracy at an early stage because of inter-class visual resemblance, different illumination, background noise, and complicated leaf texture. The study is based on the issue of strong multiclass crop disease detection in a Hybrid CNN+RNN model. This study aims at developing an effective deep learning model that would be able to extract both spatial and contextual information in order to classify reliably between ten different classes: (Maize Brown Spot, Maize Healthy, Maize Rust, Potato Early Blight, Potato Healthy, Potato Late Blight, Soybean Healthy, Soybean Mosaic Virus, Soybean Pod Mottle and Soybean SBS). The CNN block gets prejudiced spatial characteristics, and the RNN block get contextual dependencies to enhance the separability of the classes. The results of the experiments are good with Maize Brown Spot making 20 correct predictions, Potato Healthy, Soybean Mosaic Virus and Soybean Pod Mottle having 17 correct classifications and training accuracy reaching to around 85 with the loss staying between about 0.6 and 1.0. Nevertheless, the similar categories visually similar to each other including the category of (Maize Rust) and Soybean SBS revealed problems when the precision and recall exhibited imbalance to the middle. Such difficulties were alleviated by hybrid merging of features and optimization at the expense of optimization, which enhanced feature clustering as demonstrated by t-SNE visualization. In general, the suggested hybrid model increases disease discrimination potential and proves to be a good candidate of scalable and real-time precision agriculture use.

Zitieren

APA

Monika, & Kumar, A. (2026). Deep hierarchical spatial–sequential framework for robust multiclass leaf disease analytics. IET Conference Proceedings, 2026(15), 674-678. https://doi.org/10.1049/icp.2026.3973

Vancouver

Monika, Kumar A. Deep hierarchical spatial–sequential framework for robust multiclass leaf disease analytics. IET Conference Proceedings. 2026;2026(15):674-678. doi:10.1049/icp.2026.3973

BibTeX
@article{Monika2026,
  author = {Monika and Kumar, Ankit},
  title = {Deep hierarchical spatial–sequential framework for robust multiclass leaf disease analytics},
  journal = {IET Conference Proceedings},
  year = {2026},
  volume = {2026},
  number = {15},
  pages = {674--678},
  doi = {10.1049/icp.2026.3973}
}
Kostenlos starten →