Deep Learning Based Approach for Identification of Local Fish

Abstract

Bangladesh is considered one of the most suitable area for fish culture with the world's largest climate wetland with thousands of rivers and ponds and being a fish-loving nation. Learning a classification of fish can help people to identify the local fish. Various types of fish are classified based on their characteristics so that people can scientifically describe different types of fish easily. The classification of the different species of internal relationships and their consensus of an animal is specially considered. Differences between fish size and size are so much that it is very difficult to detect them. The people of new era use Mobile phones and other devices to shoot fishes but they became confused to identify fish. For this reason, we take a purpose for identifying fish. For our experiment, we have used total 6000 images of local fish with 10 categories. We have used Convolutional Neural Network models, those are Inception-V3, MobileNet, ResNet50, and Xception and they have obtained a high accuracy of 98.07%, 98.41%, 97.65%, and 95.53% respectively on our dataset.