Feature Selection for Cotton Leaf Disease Using Deep Learning

Detecting diseases that affect cotton leaves is essential for guaranteeing the health of crops and increasing yields in agricultural settings. The application of deep learning approaches has demonstrated encouraging outcomes in the process of automating illness detection; yet, feature selection continues to be an essential component for improving model performance and interpretability. Within the scope of this work, a strategy to feature selection in cotton leaf disease detection was investigated by utilizing VGG16 deep learning techniques. Early leaf diseases on cotton plants were identified over the course of this research. However, the majority of illnesses that affect cotton plants are found on the leaves rather than the root bell. The deep learning model that we developed for the purpose of identifying cotton plant leaf diseases was divided into six unique categories when it came to our work. These categories were healthy, curly virus, fusarium wilt, powdery mildew, bacterial blight, and nutrient deficient images of the cotton plant. Among the goals of feature selection are the creation of models that are easier to grasp and more straightforward, the enhancement of data mining efficiency, and the preparation of data that is clean and easy to comprehend. The strategy of feature selection has been shown to be successful in enhancing the classification accuracy of the deep learning model, as demonstrated by the results of the experiments. In addition, the selected characteristics offer insights into the essential traits that are symptomatic of cotton leaf diseases. This helps agronomists understand disease trends and makes it easier for them to implement tailored treatments for crop protection.