Fine-Grained Analysis of Birch Leafminer Disease: Varying Severity Five Class Ensemble Twin CNN Model

This research explores a cutting-edge method in the area of agricultural disease control. The method relies on an ensemble model built around Twin Convolutional Neural Networks (CNN) architecture. As part of an evaluation task, a dataset was compiled for over 90 thousand high-resolution images. It includes diseases variously manifested under different environmental conditions. The ensemble model performed extremely well, with an overall accuracy of 97.05
^%
The model's accuracy in disease classification concerning evaluation metrics such as precision, recall and F1-score was independently verified. A confusion matrix was used to analyze the model, exposing its strong points and weaknesses. High-accuracy models could prevent early outbreaks of the disease, reducing crop loss and increasing productivity. The model will be refined in future versions by cleaning up specific misclassification patterns in that confusion matrix. This iterative purification process aims to strengthen the model's effectiveness, further reinforcing its key role in promoting precision agriculture practice.