Deep Learning in Agriculture: Cotton Leaf Disease Identification Using CNNs

This research proposes a new methodology for detecting cotton leaf diseases by Convolutional Neural Networks (CNNs). The proposed system uses high-quality images of cotton leaves to identify diseases, including bacterial blight, curl virus, and fungal infection, with minimum human intervention. Thus, the study concerns dataset preparation, model architecture design, training, and evaluation. This CNN model was trained at an accuracy of 95.04% and validation accuracy of 84.38%, which suggests it can be used for disease detection. The test dataset evaluation's precision, recall, and F1 scores were reported as 80%, 79%, and 79%, respectively. Finally, these results demonstrate the potential of CNNs in process automation for diagnostic purposes and their contribution to precision agriculture. Future work could reduce model generalization, use regularization techniques, and increase the dataset's size.