Comparative Study Between Transfer Learning Models to Detect Shrimp Diseases

Shrimps are one of the most important animals in aquaculture. Over the past fifty years, there has been a steady increase in shrimp production worldwide. Shrimp production reached 5.5 tonnes in 2021, and that many countries tend to increase their CAGR and production. Some major problems and challenges persist in shrimp production, such as feed quality and availability, production cost, seed quality, and diseases. There are types of diseases such as black gill and white spot disease. Any delay in the detection of the diseases can lead to the loss of shrimp and infection of other shrimp. In this paper, the authors used transfer learning models to detect two types of shrimp disease (white spot disease and black gill) and to detect diseased shrimp from normal shrimp. the authors aim to know the best transfer learning model that has the highest accuracy in the early detection of shrimp disease. Using five types of transfer learning, the model with the highest validation accuracy is MobileNetV1, with 95% in experiment one and 92.5% in experiment two.