TOWARDS IMPROVED CLASSIFICATION OF POTATO DISEASE BASED ON HYBRIDIZATION OF ResNet ARCHITECTURE AND NAIVE BAYES CLASSIFIER

Crop diseases are a major challenge in agriculture as they affect farm produce and often lead to reduced yields thereby threatening food security. Several studies have proposed different Machine Learning (ML)-based approaches for the classification of diseases in crops. In this paper, the focus is on building two-stage Potato disease classification model using Residual Network (ResNet-18) architecture and Naive Bayes (NB) algorithm. The image crop dataset was collected from Mendeley repository. The Potato disease classification model was built by using Resnet-18 architecture at feature extraction layer while Gaussian Naive Bayes was used for the potato image classification task. The crop classification model was trained and tested on the chosen dataset of potato crop images using training and test ratio of 0.80 and 0.20 respectively. Experimental results showed that the use of the ResNet-18 model helped the NB-based potato disease classification model to achieve high accuracy in classifying both healthy and diseased crops. The model achieved 94.45%. The study demonstrated how the hybridization of a deep learning approach for feature selection and a classical learning algorithm called Naive Bayes for the potato crop disease classification. Thus, the proposed model can be effective for potato disease classification there reduce losses, and improve agricultural productivity.