Deep learning with convolutional neural networks to assess rice plant diseases performs very well

08 Jul 2025

UP Manila

As with many Asian countries, rice is a principal food in the Philippines providing nearly half of the daily caloric needs of Filipinos. However, rice plants are also susceptible to many diseases whose spread is induced by weather conditions such as high humidity and rainfall producing detrimental effects on crop’s yield and thus affecting the country’s food security. Hence, a rapid, early, and correct detection of rice plant disease is crucial to prevent spread of the disease mitigating its detrimental effects through an early institution of preventive measures.

In this study, researchers applied a deep learning approach using convolutional neural networks in the assessment of rice plant disease. Results showed superior diagnostic performance with these models. As such, these deep learning models can be useful complementary tools which can be deployed as quick and non-invasive diagnostic support instruments assisting farmers in the evaluation of rice diseases especially in communities where agricultural experts are limited. Farmers would gain more valid outcomes with these new technological diagnostic approaches, enabling them to institute cost-effective measures. Thus, effective management of rice plant diseases, optimized and efficient use of available resources leading to improved rice crop productivity can be achieved. A working partnership of agriculturists and machine learning enthusiasts is crucial to achieve the desired goal of early identification of rice plant diseases for prompt intervention efforts to be instituted.

The purpose of the study is to ascertain the distinguishing capability of convolutional neural networks in the recognition of rice plant disease. Deep learning models (base convolutional neural and pre-trained networks) were applied to the Philippine Rice Disease Dataset to diagnose rice plant diseases. VGG16 obtained the best performance with a 96% accuracy, 99% sensitivity, 97% precision, 98% F1-score, and a 0.834 normalized Matthews Correlation Coefficient. InceptionV3 also generated superior performance while the base model had a lower diagnostic capability. These models can be useful complementary tools which may be deployed as quick and non-invasive diagnostic support instruments assisting farmers in the evaluation of rice diseases especially in communities where agricultural experts are limited. Farmers would gain more valid outcomes with these technological approaches, enabling them to institute cost-effective measures. Thus, effective management of rice plant diseases, optimized and efficient use of available resources leading to improved rice crop productivity can be achieved.

Authors: Vincent Peter C. Magboo and Ma. Sheila A. Magboo (Dept. of Physical Sciences and Mathematics, University of the Philippines Manila)

Read the full paper: https://ieeexplore.ieee.org/document/102867499

Deep learning with convolutional neural networks to assess rice plant diseases performs very well

As with many Asian countries, rice is a principal food in the Philippines providing nearly half of the daily caloric needs of Filipinos. However, rice plants are also susceptible to many diseases whose spread is induced by weather conditions such as high humidity and rainfall producing detrimental effects on crop’s yield and thus affecting the country’s food security. Hence, a rapid, early, and correct detection of rice plant disease is crucial to prevent spread of the disease mitigating its detrimental effects through an early institution of preventive measures.

In this study, researchers applied a deep learning approach using convolutional neural networks in the assessment of rice plant disease. Results showed superior diagnostic performance with these models. As such, these deep learning models can be useful complementary tools which can be deployed as quick and non-invasive diagnostic support instruments assisting farmers in the evaluation of rice diseases especially in communities where agricultural experts are limited. Farmers would gain more valid outcomes with these new technological diagnostic approaches, enabling them to institute cost-effective measures. Thus, effective management of rice plant diseases, optimized and efficient use of available resources leading to improved rice crop productivity can be achieved. A working partnership of agriculturists and machine learning enthusiasts is crucial to achieve the desired goal of early identification of rice plant diseases for prompt intervention efforts to be instituted.

The purpose of the study is to ascertain the distinguishing capability of convolutional neural networks in the recognition of rice plant disease. Deep learning models (base convolutional neural and pre-trained networks) were applied to the Philippine Rice Disease Dataset to diagnose rice plant diseases. VGG16 obtained the best performance with a 96% accuracy, 99% sensitivity, 97% precision, 98% F1-score, and a 0.834 normalized Matthews Correlation Coefficient. InceptionV3 also generated superior performance while the base model had a lower diagnostic capability. These models can be useful complementary tools which may be deployed as quick and non-invasive diagnostic support instruments assisting farmers in the evaluation of rice diseases especially in communities where agricultural experts are limited. Farmers would gain more valid outcomes with these technological approaches, enabling them to institute cost-effective measures. Thus, effective management of rice plant diseases, optimized and efficient use of available resources leading to improved rice crop productivity can be achieved.

Authors: Vincent Peter C. Magboo and Ma. Sheila A. Magboo (Dept. of Physical Sciences and Mathematics, University of the Philippines Manila)

Read the full paper: https://ieeexplore.ieee.org/document/102867499