Ensemble Learning in Rice Leaf Diseases Classification

Rice leaf diseases are a major threat to agriculture, impacting global food security and economic stability. Detecting these diseases early and accurately is essential to reducing crop losses and increasing yields. In this study, we have explored how well deep learning models can identify and classify leaf diseases using images. We have utilized State-of-the-art (SOTA) Convolutional Neural Network (CNN) models, including ResNet50 and DenseNet169, pre-trained on ImageNet dataset as feature extractors. The extracted features are then classified using linear classifiers, namely, Logistic Regression, Linear Support Vector Machine (LSVM), Perceptron, and combination of Logistic Regression, Linear Support Vector Machine (LSVM), Perceptron, quadratically penalized SVM, and SVM with quadratically smooth loss utilizing an ensembling approach. The results show that applying the ensembling approach to classify the features extracted using ResNet50 performs the best, achieving 96.45% accuracy, 96.48% precision, 96.45% recall, and 96.43% fl-score. This study highlights the potential of an ensembling approach on a combination of Deep Learning (DL) models with Transfer learning (TL) and Machine Learning (ML) classifiers to provide reliable and scalable tools for identifying rice leaf diseases, which could be used to help farmers detect issues with rice leafs and take timely action.