An AI-Powered Diagnose of Plant Disease Using CNN

The inclusion of Artificial Intelligence (AI) in precision agriculture is growing rapidly, particularly in the automated detection of plant diseases. Conventional CNN-based models have shown success. However, they often fail to focus attention on the most significant parts of images, where vital information about the disease is found. This work proposes an extended CNN architecture that is integrated with Convolutional Block Attention Module (CBAM) for enhancing channelwise and spatial-wise feature focus in plant disease detection in tomato, potato, and bell pepper leaves. The proposed pipeline consists of preprocessing, segmentation, data augmentation, feature extraction, and classification, and is tested on the PlantVillage dataset with six classes of diseases/health. The CNN-CBAM model attains an overall accuracy of around 98.0% with increased precision, recall, and F1-score compared to a baseline CNN and a MobileNet-based approach, which makes it robust and capable of generalization. The key contributions are multicrop attention-driven architecture, end-to-end pipeline, and comparative analysis of the proposed method with the state-of-the-art deep learning methods for plant disease diagnosis.