Automated Detection of Multiple Plant Diseases in Paddy and Jackfruit using Machine Learning

Abstract

Early identification of plant diseases plays an important role in reducing crop damage and improving agricultural productivity. This study presents a deep learning–based system for the automatic detection of multiple diseases affecting paddy (rice) and jackfruit plants using images of leaves and fruits. The dataset used in this research was developed using publicly available sources along with images collected from nearby farms. It consists of nine balanced disease classes, including Healthy Rice Leaf, Healthy Jackfruit Leaf, Rice Leaf Blast, Rice Leaf Scald, Rice Leaf Tun grow, Rice Leaf Insect Damage, Rice Rot Disease, Black Spot Disease in Jackfruit Leaves, and Jackfruit Fruit Diseases, with a total of around 2,100 images. Before training the models, all images were pre-processed by resizing, normalization, noise reduction using Gaussian filtering, and data augmentation techniques to improve learning performance. Two deep learning approaches were implemented for comparison: a custom Convolutional Neural Network (CNN) and a transfer learning model based on MobileNetV2 with fine-tuning. The models were evaluated using validation accuracy, top-2 accuracy, and validation loss. The experimental results show that the MobileNetV2 model performs significantly better than the basic CNN model, achieving a validation accuracy of 84.94%, a top-2 accuracy of 90.78%, and a validation loss of 0.9879. These findings highlight the effectiveness of transfer learning for plant disease detection, especially when working with balanced datasets of moderate size. The proposed system offers a practical and affordable solution that can help farmers detect plant diseases at an early stage and take timely preventive actions.