Coffee Leaf Disease Classification by Using a Hybrid Deep Convolution Neural Network

The most common symptoms of coffee leaf disease are coffee leaf rust, black rot diseases, and brown eye spot. Leaf rust is the first symptom of coffee leaves that is visible on the upper surface of the leaves, i.e., small pale-yellow spots. At later stages, these spots gradually expand their diameter and become visible on the undersurface. The underside of the leaves becomes orange–yellow in colour or red–orange powdery, which varies from region to region. Finally, these leaf diseases rapidly spread and result in significant economic losses, so it's essential to identify plant diseases in their initial phases. Detailed features of the leaves have been required for this task to differentiate healthy and unhealthy coffee leaves. The proposed model utilized a mix of two deep learning methods, i.e., InceptionV3 and DenseNet121, and performed better using the concatenation and augmentation approach of convolution neural network (CNN). The proposed work contrasts existing pre-trained CNN models, such as InceptionV3 and DenseNet121, which require more computing power and parameters. According to the proposal, the hybrid deep convolution neural network model demonstrated a remarkable accuracy rate of 99%, which proved better than existing models such as InceptionV3, EfficientNetB0 and DenseNet121, which gained 94%, 98% and 96.36% accuracy, respectively.