EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Federated Learning-Based Potato Leaf Disease Detection in Uncontrolled Agricultural Environments

Akshita Patwal, Darshan Singh, Ishita Patwal

Procedia Computer Science · 2026 · 283 · 4050-4060

DOI: 10.1016/j.procs.2026.06.466 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Procedia Computer Science, 2026. Verfasst von Akshita Patwal und 2 weiteren Autor:innen. Die Arbeit nennt 17 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

One of the most important technology solutions to address the problems of climate change, food security, and sustainable agriculture is smart farming. Yield losses may be avoided in large part by automated agricultural disease identification, however conventional centralized learning techniques are hampered by privacy issues, bandwidth constraints, and farmers’ unwillingness to disclose sensitive data. Here we use the Potato Leaf Disease Dataset in Uncontrolled Environment, which comprises 3,076 photos from seven classes (phytophthora, bacterial, fungal, viral, nematode, pest, and healthy) to present a federated learning framework for potato leaf disease detection in this research. The challenge is made further difficult by the dataset’s accurate representation of agricultural circumstances, which includes fluctuations in light, background noise, and leaf orientation. Our approach compares IID and non-IID client distributions using three CNN backbones (ResNet50, MobileNetV2, and EfficientNet-B0) in a federated setting. FedProx is integrated to handle statistical heterogeneity. According to experimental data, FedProx raises non-IID performance from 96.2% to 97.2%, while the federated ResNet50 model attains 98.6% accuracy, closely matching the centralized baseline (99.1%). Convergence research also shows that, in comparison to their lightweight equivalents, deeper models like ResNet50 stabilize more quickly (30 rounds). These findings establish the viability of federated learning in real-world agricultural environments and show its promise for scalable, privacy-preserving smart farming applications.

Zitieren

APA

Patwal, A., Singh, D., & Patwal, I. (2026). Federated Learning-Based Potato Leaf Disease Detection in Uncontrolled Agricultural Environments. Procedia Computer Science, 283, 4050-4060. https://doi.org/10.1016/j.procs.2026.06.466

Vancouver

Patwal A, Singh D, Patwal I. Federated Learning-Based Potato Leaf Disease Detection in Uncontrolled Agricultural Environments. Procedia Computer Science. 2026;283:4050-4060. doi:10.1016/j.procs.2026.06.466

BibTeX
@article{Patwal2026,
  author = {Patwal, Akshita and Singh, Darshan and Patwal, Ishita},
  title = {Federated Learning-Based Potato Leaf Disease Detection in Uncontrolled Agricultural Environments},
  journal = {Procedia Computer Science},
  year = {2026},
  volume = {283},
  pages = {4050--4060},
  doi = {10.1016/j.procs.2026.06.466}
}
Kostenlos starten →