Deep Transfer Learning with GUI for Rice Pests Classification and Preventive Recommendations

Rice serves as a crucial crop for the global food security as it inhabits a core staple among billions of people in the world. However, rice yield is increasingly threatened by pest infestations, which can cause severe economic and production losses. Therefore, early and accurate identification of rice pests is important for timely measures against the worsening losses. Deep learning technology has proven to be an efficient solution to pest classification, which indeed gives exceptional accuracy in classifying images. However, these deep learning models require high computational and memory power, and this study leverages transfer learning to address those problems from four pre-trained CNN models: ResNet50, InceptionV3, VGG19, and DenseNet201 for identifying rice pests. A publicly accessible rice pest dataset was used in this endeavor and intensive tests were carried out for performance evaluation of the models identified. Among the models evaluated, DenseNet201 perform well with the classification accuracy of 86% that clearly indicates its potentiality in accurate pest identification. An easy and user-friendly GUI was created through which this technology reaches farmers and allows them to upload pest images for real-time classification and identification of pest species with customized preventive actions. The integration of transfer learning and GUI thus provides an affordable, scalable, and easy to use tool for early pest management.