An Explainable Approach for Detecting Potato Leaf Disease Using Ensemble Model

The potato is one of the most significant crops in the world, yet it is susceptible to a number of diseases, especially early and late blight. Convolutional neural networks (CNNs) are one deep learning method that has been utilised recently to forecast plant diseases from images. In this study, we explored the use of two well-known CNN architectures, Inception v3 and ResNet152 V2, for disease prediction in potato leaves. Images of leaves in various stages of early blight, late blight, and healthy made up our dataset. On this dataset, we trained and evaluated the CNN models, and we were able to achieve high accuracy for Inception V3 and ResNet152 V2 of 98.49% and 98.78%, respectively. The accuracy was then increased to 99.54% by combining the two models into an ensemble model. Our results show the potential of deep learning approaches for precise and effective prediction of potato leaf diseases, which can assist farmers in promptly taking preventative measures to avoid crop losses.