Abstract:
Fish disease is one of the major problems in the field of fish farming in asian region.
Every year farmers have to face a lot of losses in their business for the fish disease.
Especially EUS (Epizootic Ulcerative Syndrome) is one of the worst disease by which
they are most commonly affected. It is very difficult for them to identify fish disease
because they don‘t have enough knowledge about fish disease. So in this paper we
actually try to help them to figure out this problem. Here we try to build a model which
can automatically detect if it is a EUS disease or Non EUS disease. In earlier a few
researches has been conducted to identify fish disease using machine learning but in
those research there were lacking‘s of enough authentic data and lower accuracy. In this
paper, I have proposed a machine learning approach using InceptionV3 to detect EUS
and Non EUS disease with an accuracy of 95.74%. I have used total 938 images of data.
Among 80% of the data used for training purpose and 20% used for testing purpose. For
the experiment, I have also used some other pre-trained models for example VGG16,
Xception, MobileNetV2 and InceptionRestNetV2 to find the best model for my project.
Here we use data augmentation technique to enhance the images and increase the
accuracy. The proposed combination of Transfer Learning and Data Augmentation
techniques gives better accuracy as compared to the others.