Plant Disease Detection

Agriculture remains a cornerstone of the global economy, yet plant diseases continue to threaten crop yields and food security. Traditional disease identification methods rely on manual inspection, which is often labor-intensive, inaccurate, and impractical for large-scale farming. This project introduces a deep learning solution that utilizes Convolutional Neural Networks (CNNs) to automate the classification of 38 plant diseases based on leaf images, significantly improving the efficiency and accuracy of disease diagnosis.

To train the model, we utilized the PlantVillage dataset, containing over 55,000 images of healthy and diseased plant leaves. The images were resized to 128x128 pixels, normalized, and augmented using flipping, cropping, and brightness adjustments to enhance generalization. The CNN model architecture consisted of three convolutional layers (32, 64, 128 filters), pooling layers, a fully connected dense layer (128 units), and a dropout layer (0.5) to prevent overfitting. The model was optimized using Adam optimizer and trained on a TPU v2–8, significantly improving processing speed. The system achieved 91% test accuracy and was externally validated on unseen images, obtaining 92% classification confidence.

This project highlights the transformative potential of AI in agriculture, enabling farmers to detect diseases early and take preventive measures. Future improvements include expanding the dataset with real-world farm images, implementing transfer learning for improved adaptability, and deploying the model as a mobile or web-based diagnostic tool. With continued enhancements, this AI-powered approach can revolutionize plant disease management, contributing to sustainable farming and global food security.