A Cross-Domain Hybrid Model for Paddy Disease Classification

Paddy cultivation in the nation has been a way of life for many centuries. Farmers face a formidable challenge in identifying these paddy diseases early and accurately to prevent significant yield losses. Therefore, this paper proposes a cross-domain model for detection of paddy diseases using deep learning and machine learning algorithms. In order to categorize 13 types of paddy diseases, the proposed hybrid model was developed. For this task, we dispensed with the need for a custom-built CNN architecture and instead used a pre-trained one (VGG16) in conjunction with four machine learning algorithms. In this context, CNN's 4096 high-level features of 20,800 images are classified using machine learning algorithms rather than a fully connected layer for classification. Naive Bayes has an average accuracy of 99.9% in the hybrid model. Due to the large number of images required for training the CNN model, this hybrid model is being hand-picked. Therefore, it is possible that the proposed method is an effective method for detecting the paddy disease.