Manuscript received July 24, 2025; revised August 19, 2025; accepted October 10, 2025; published February 27, 2026.

Abstract—Plant diseases are one of the most significant threats to global food security, impacting food availability and safety, as well as their negative impact on agricultural productivity. Addressing this major challenge requires the development of advanced and modern disease diagnostic equipment that not only detects but also accurately identifies disease. In this paper, we propose an innovative algorithm based on transfer learning for the detection of tomato leaf diseases. Convolutional Neural Networks (CNNs) were used in conjunction with pre-trained models such as EfficientNetB3, Xception, and MobileNetV2 to accurately classify and diagnose diseases. Experimental results demonstrated high classification accuracy, reaching 0.93 for CNN, 0.94 for MobileNetV2, 0.95 for Xception, and 0.995 for EfficientNetB3. Using machine learning and image processing techniques, the system can quickly and automatically diagnose disease symptoms on plant leaves, providing farmers with up-to-date and reliable information on what to do. The research also addresses the interpretability of the model by applying Saliency Maps and Grad-CAM (Gradient-Weighted Class Activation Mapping) techniques to visually explain the model’s decisions, which enhances transparency and increases user confidence in the Artificial Intelligence (AI)-based system in the agricultural field.

Keywords—machine learning techniques, convolutional neural network, image diagnostic, saliency maps, Grad-CAM, explainable AI, leaf disease detection

Cite: Youssef Laatiri and Mohamed Ali Mahjoub, "Explainable Intelligent Detection of Tomato Leaf Diseases: An Explainable Deep Learning Approach," Journal of Image and Graphics, Vol. 14, No. 1, pp. 141-148, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.