162
BULLETIN OFFICIEL
Nº 7366 – 1er rejeb 1446 (2-1-2025)
Considered diseases The following image displays the 10 various diseases that our application will detect.
Specifically, the 7176 dataset's images are imbalanced in distribution across the 10 diseases. The considered number of images per disease is outlined in Table 1. A portion of this dataset is devoted to training, while the remainder is for testing. It is worthwhile to mention that the learning rate has been adjusted in order to enhance the overall accuracy. Nevertheless, the training step duration takes longer. Though, the choice of such a parameter is crucial to maintain a certain trade-off between accuracy and time complexity.
Results: Several optimization methods have been employed including (i) Stochastic gradient descent, (ii) Adagrad, (iii) Adam, (iv) Proximal Gradient Descent, and (v) Proximal Adagrad. The obtained accuracy varies between 88.3% and 89.2%. This latter is corresponding to the Proximal Gradient Descent algorithm. References [1] W. Tan, C. Zhao, and H. Wu, “Intelligent alerting for fruit-melon lesion image based on momentum deep learning,” Multimedia Tools and Applications, vol. 75, no. 24, pp. 16741-16761, 2016. [2] E. Fujita, Y. Kawasaki, H. Uga, S. Kagiwada, and H. Iyatomi, “Basic investigation on a robust and practical plant diagnostic system,” in 2016 15th IEEE Intern. Conf. Machine Learning and Applications (ICMLA), 2016, pp. 989– 992. [3] S. P. Mohanty, D. P. Hughes, and M. Salathe, “Using deep learning for image-based plant disease detection,” Frontiers in Plant Science, vol. 7, 2016. Le texte en langue arabe a été publié dans l’édition générale du « Bulletin officiel » n° 7365 du 28 joumada II 1446 (30 décembre 2024).
46