OPDD: Optimized Deep CNN Architecture for Accurate Plant Disease Detection

Plant disease detection through image classification is critical for improving agricultural productivity. This paper introduces an innovative approach for identifying plant diseases through an enhanced Convolutional Neural Network (CNN) model. The model discussed here is trained on a distinctive collection of data comprising images of plant leaves, each showcasing various plant diseases. Our architecture features multiple convolutional layers, increased neurons, and a fine-tuned learning rate of 0.0001-lower than the standard 0.001-to prevent overshooting during training. Dropout layers are also employed to minimize overfitting. The proposed approach was comprehensively evaluated using multiple metrics, including precision, recall, F1-score, and AUC. With an accuracy of 0.9963, this method outperforms current ones, leading to better plant disease detection and improved farming methods.