This research has developed a plant disease diagnosis system based on convolutional neural networks (CNN) for precision agriculture. The trained CNN is built on a data set of more than 87,000 labeled images under 38 classes and observed over 95% accuracy for detecting diseases using the texture, lesions, and colors of the leaves as effective features. Data preprocessing techniques such as augmentation and regularization have improved robustness, while the Adam optimizer ensured efficient weight updates. The system proposes scalable and automated solutions for early intervention to prevent losses. The implementation details are subsequently followed by future directions to improve performance.