An Explainable Dual-Stream Center-Weighted Smart Zoom Architecture for Robust Potato Leaf Disease Classification

DOI

10.2991/978-94-6239-799-6_8How to use a DOI?

Keywords

Potato Disease Detection; Deep Learning; EfficientNetV2B0; Attention Mechanism; Smart Zoom; Explainable AI; Cross-Dataset Generalization

Abstract

Accurate identification of potato leaf diseases is very important for reducing potato crop loss. Deep learning models get excellent accuracy on lab controlled datasets but they do not achieve the same accuracy on real-world datasets, because they have complex background. Most existing methods use information from the entire leaf and try to predict the result. But they do not pay sufficient attention to the diseased portion of the leaf.

This paper presents a Dual-Stream Center-Weighted Smart Zoom Architecture for strong potato leaf disease classification. This architecture combines a Center-Weighted Multi-Scale Attention Module with a Gaussian center prior to highlight disease-affected regions and ignore background. This attention map guides to a weakly supervised Smart Zoom mechanism that automatically extracts disease related region (Region of interest) from the full image. This extracted ROI is analyzed with the original image using a shared EfficientNetV2B0 backbone in a dual-stream architecture. A SHA-256-based dataset preparation mechanism removes duplicate images and verifies that there is no data leakage in training, validation, and testing subsets.

This proposed architecture was tested on the PlantVillage, PLD Pakistan, and PLD Indonesia datasets. On the challenging PLD Indonesia dataset, it achieved an overall accuracy 88.12%, a balanced accuracy 88.25%, and an MCC 0.8555. For model reliability, Grad-CAM visualizations are integrated in this architecture, so that we can visualize that model focuses on diseased spot or not. Overall, the proposed architecture provides an accurate and reliable solution for automated potato leaf disease classification.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Hitendra Singh
AU  - Ajay Kumar Sharma
AU  - Mayank Patel
PY  - 2026
DA  - 2026/09/30
TI  - An Explainable Dual-Stream Center-Weighted Smart Zoom Architecture for Robust Potato Leaf Disease Classification
BT  - Proceedings of the International Conference on Emerging Trends and Technologies in Applications of Computer Science (ICETTACS 2026)
PB  - Atlantis Press
SP  - 104
EP  - 120
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-799-6_8
DO  - 10.2991/978-94-6239-799-6_8
ID  - Singh2026
ER  -