Diseased Shrimp Detection based on Transfer Learning

Abstract

Shrimp farming plays an important role in aquaculture in the central coastal provinces of Vietnam. Shrimp disease is a significant threat to nutritional security and causes considerable economic loss. Identification of infected shrimps in aquaculture remains challenging due to the dearth of necessary infrastructure. The identification timely is an obligatory step to thwart from spread of disease. This paper proposes a technique to detect shrimp diseases based on transfer learning. This work includes three main steps. The first step collects and preprocess the image dataset of diseased shrimp collected from shrimp farms in Quang Nam province. The second step trains the dataset through three models SVM, VGG16, and the proposed model GonCNN. The third step tests and evaluates the accuracy of these models. Experimental results show that GonCNN has an accuracy of up to 92.93%, while SVM and VGG16 with 75.67% and 86.94% of accuracy, respectively.