A Fine-tuned Deep Learning-based VGG16 Model for Cotton Leaf Disease Classification

Cotton is called "white gold" which can be grown in countries like India and China, the most. The production of cotton can be degraded by various diseases affecting their leaves and the overall health of the plant. The proposed work has been presented for pioneering research into cotton leaf disease detection, proving the power of the VGG16 model. Deep learning methods have made an important contribution to obtaining remarkable accuracy of 95.5 %. Different batch size such as 8, 16, 32, and 64 have been set for analyzing the performance among which the highest and lowest performance has been shown by batch size values 8 and 64, respectively. Additionally, the testing loss value became a minimum of 0.0253 at epoch 32, indicating the proposed approach is the optimal and most effective for classifying healthy cotton leaves from diseased ones. This study involves a comprehensive image analysis of the various conditions and stages of development. The fine-tuned VGG16 model demonstrated remarkable skill in recognizing deep patterns denoting the cotton leaf health condition. As the dataset contained a limited number of images, hence, the two different augmented operations on each image have been applied resulting in 5000 cotton leaves images. This augmentation strategy was crucial, allowing VGG16 to break through with an accuracy of 95.5% and become the best model at classifying cotton leaf diseases. Besides its value in predicting the state of cotton plants, this research illuminates deep learning's revolutionary potential across diverse fields. The proposed approach has laid the groundwork for incorporating such models into applications dealing with the classification of other plant diseases.