Transfer Learning Approach for Epizootic Ulcerative Syndrome and Ichthyophthirius Disease Classification in Fish Species

A substantial contributor to the national economy in terms of food, nutrition and economic progress is the Indian Fish industry. Along with satisfying domestic demand, this sector has significantly increased its contribution to foreign exchange revenue by export. The fish industry is currently developing in a direction that will increase the commercialization and specialization of aquatic species. The major bottleneck in the development of fish industry is fish diseases as it affects the productivity rate heavily. Thus, it is essential to predict fish disease at early stage. In this paper the computer vision technique is implemented for automatic detection and classification of fish disease in fish industry. A transfer learning strategy is proposed for automatic fish disease detection. Healthy fish, fish with EUS disease, and fish with Ichthyophthirius disease are the three classes considered for the analysis. The dataset is trained by proposed VGG16 model and is compared with AlexNet. Low level features are better extracted using VGG16 which plays essential role in classification. AlexNet provides accuracy of 64%, while VGG16 produces accuracy of 91%.