SDNet: Integrated Unsupervised Learning with DLCNN for Shrimp Disease Detection and Classification

Abstract

Shrimp is a main international food item with a significant economic value, as well as one of the most vital animal protein sources. However, the production of shrimps is directly affected by the different types of shrimp diseases. Thus, it is necessary to identify the shrimp diseases in primary stage to avoid the losses. Therefore, this article is implemented the shrimp disease network (SDNet) using deep learning architectures. Initially, K-means clustering (KMC) is applied on the test images to localize the region of disease or virus location. Then, machine learning based iterative random forest algorithm (IRFA) is utilized to extract the features from segmented images and it also develops the optimal features. Finally, deep learning convolution neural network (DLCNN) is used to perform the multi class classification of shrimp diseases by training the optimal features. The proposed SDNet method resulted in superior performance as compared to state of art approaches with respect to both subjective and objective metrics in terms of classification metrics such as sensitivity, specificity, accuracy, precision, recall, and F1-socre.