Deep Learning-Based Classification of Rice Leaf Diseases Using Hybrid Ensemble Models

This research aimed to identify the possibility of applying deep learning to rice leaf diseases diagnostic procedures. The study applied different data preprocessing techniques that were the key to getting the data ready for the analysis process to a set of commonly used rice leaf images. The first process implemented was training different kinds of deep learning models, such as ResNet and DenseNet architectures, on the preprocessed data. To ensure the best possible performance, the top single models from each set were then combined into all the ensemble methods in a way that made them the most potent. The pre-trained models used were ResNet152V2, DenseNet121, InceptionResNetV2, MobilityNetv2, and all models were trained on the preprocessed data. The algorithms' success in classification was assessed based on the measures of accuracy, precision, recall, and F1 score. The most successful classifiers were combined into hybrid ensembles using the Ensemble method. The holistic approach to this matter brought the desired result that is a 98% accuracy in the classification of rice leaf diseases, which is truly a remarkable result that no single model can bring.