Enhancing Paddy Leaf Disease Recognition and Diagnosis with Convolutional Neural Networks and Ensemble Transfer Learning

Early detection of crop diseases is an opportunity to optimize agricultural care and ultimately increase crop yields. Researchers studying machine learning are becoming increasingly interested in the automated detection of diseases affecting paddy leaves in photos. Methods rely on convolutional neural networks are extensively used to successfully tackle these challenges. In our current study, we utilize well-known transfer learning models, including, MobileNetV2, Xception, DenseNet201, and VGG19, to diagnose rice plant diseases using publicly available Kaggle dataset.To elevate the diagnostic system’s performance, we introduce the concept of ensembling transfer learning (TL) models in this research. We explore all feasible combinations of TL models and conduct experiments with each ensemble. Importantly, several of the ensemble models surpass the individual TL models in terms of performance. This study underscores that specific ensemble models exhibit superior performance compared to advanced learning models like Convolution-Boosting algorithms.