Classification of Underwater Sea Species Using Convolutional Neural Network

Authors:
Publisher website
Google Scholar
Abstract
The importance of recognizing fish species has increased as a result of new marine science research. Classifying fish species manually requires a person who has knowledge about all the species and this is a time consuming process. The classification of fish species could be automated using technology, advancing marine science. Deep learning techniques have led to an increase in the number of image classification tasks. In this paper we have compared performance of proposed Convolutional Neural Network (CNN) model with existing CNN architectures for classification of LifeClef 2015 Fish (LCF-15) dataset. This dataset consist of 15 different classes of sea species and total number of images are 22444. It is observed that the proposed model achieves accuracy of 98.76%. The proposed model is light weight, having less number of parameters and flops compared to other existing models. The precision, recall and f1-score achieved for the proposed model are 98.78%,98.76% and 98.71 % respectively.

This publication has 19 references indexed in Scilit: