An Ensemble Technique Using Genetic Algorithm and Deep Learning for the Prediction of Rice Diseases
Sunanda Das, Tanvir Habib Sardar, D. S. Sahana
Abstract
The word Dhana, which means “paddy” or “rice” and also is mentioned in our Veda, was used as food and offered to the gods. But now, rice production in India is significantly affected due to three disease classes: leaf smut, brown spot, and bacterial leaf blight. A standard model for plant disease prediction can make it easier to take the necessary steps to stop losses ahead of time. The proposed model is approached here to find a way to give corrective alerts before the occurrence of any type of rice disease. In this work, some state-of-the-art classifier models are considered, and then training is performed with a dataset of these prevailing diseases in rice. The genetic algorithm-based ensemble architecture is used in a proposed method to get a better validation of the accuracy. A validation accuracy of up to 94.47% and 89.34% has been observed during testing with the InceptionV3 ResNet152V2 model and ResNet50 InceptionV3 model. The result analysis report shows the method to be a practical approach for predicting rice disease, which gives better results.