Accurate Rice Disease Detection Using Hybrid Convolutional Neural Networks and Transformer Models

Awad, Waleed Kareem; Mahdi, Eman Turki; Nafea, Ahmed Adil

doi:10.24271/psr.2025.490265.1825

Accurate Rice Disease Detection Using Hybrid Convolutional Neural Networks and Transformer Models

Document Type : Original Article

Authors

Waleed Kareem Awad ¹

Eman Turki Mahdi ²

Ahmed Adil Nafea ³

¹ Department of Computer Science, College of Computer Science and IT, University of Anbar, Ramadi, Iraq.

² Department of Computer Networks System, College of Computer Science and IT, University of Anbar, Ramadi, Iraq.

³ Department of Artificial Intelligence, College of Computer Science and IT, University of Anbar, Ramadi, Iraq.

10.24271/psr.2025.490265.1825

Abstract

Rice cultivation is significantly affected by different diseases that can severely reduce crop quality and yield. Farmers, particularly those lacking specialised knowledge, often struggle to identify and manage these diseases effectively. This study aims to create an automated system for the early detection and classification of rice diseases utilizing deep learning techniques to detect three prevalent diseases, leaf smut, brown spot, and bacterial blight; this study utilized a Convolutional Neural Network (CNN) model that had been trained on an extensive data set of photos of rice plants. The CNN model showed promise as a dependable instrument for automated disease classification by achieving a high 95% accuracy in differentiating between these illnesses. Based on the results, a system like this could help farmers manage diseases in a timely manner, improving crop results and save losses from incorrect diagnoses or postponed treatment.

Keywords

Convolutional Neural Network (CNN)

Rice Disease Classification

Deep Learning

Crop Diseases

and Image Recognition