Automated Detection of Eggplant Diseases Using Custom Convolutional Neural Network (CNN) and Transfer Learning Models

Abstract

Eggplant is a crop with substantial economic value; it is at risk for several illnesses, including small leaf, mosaic virus, bacterial wilt, and leaf spot. Timely and accurate identification of these illnesses is essential for efficient handling and increased crop productivity. This project aims to use transfer learning and convolutional neural networks (CNN) to create an automated system for diagnosing and categorizing eggplant illnesses. A 4-6 convolutional layer and 2-3 classification layer bespoke CNN model was constructed, and its performance was assessed against three pre-trained models: VGG16, InceptionV3, and MobileNetV2. We carefully adjusted the models using optimizers, epochs, and other hyperparameters to determine the best method for identifying eggplant illness. This showed that the custom CNN outperformed all models applied in this study and yielded 96.09%, which refers to the capacity of the deep learning model to classify plant diseases correctly.