Simulation Techniques of Digital Twin in Real-Time Applications

10Precision Agriculture: An Augmented Datasets and CNN Model-Based Approach to Diagnose Diseases in Fruits and Vegetable Crops

Sparsh Mehta^*, Gurwinder Singh and Yogiraj Anil Bhale

Department of AIT-CSE, Chandigarh University, Punjab, Mohali, India

Abstract

Pests and diseases pose significant threats to the fruit and vegetable industry, leading to yield reductions, compromised produce quality, and financial losses for growers. An accurate diagnosis of diseases is challenging due to similar symptoms and the possibility of multiple diseases occurring simultaneously. This study proposes a deep learning-based approach using a convolutional neural network (CNN) model to classify and identify diseases in fruits and vegetables by modeling various scenarios and treatments in the digital twin environment. This enables resource optimization and the creation of more potent disease control strategies. The approach involves creating a dataset of disease images, augmenting it with scaling and rotation techniques, and training the CNN model for disease recognition. The proposed model achieves high values for evaluation metrics such as accuracy, precision, recall, specificity, and F1-score, ranging from 96.85% to 99.3%. It effectively identifies individuals with diseases and healthy leaves, outperforming other models such as InceptionNet and EfficientNetB0. However, accurately diagnosing Marssonina leaf blotch (MLB) remains a challenge due to similarities with other diseases. Additionally, ...