Automatic Fish Species Classification Using Deep Convolutional Neural Networks

In this paper, we presented an automated system for identification and classification of fish species. It helps the marine biologists to have greater understanding of the fish species and their habitats. The proposed model is based on deep convolutional neural networks. It uses a reduced version of AlexNet model comprises of four convolutional layers and two fully connected layers. A comparison is presented against the other deep learning models such as AlexNet and VGGNet. The four parameters are considered that is number of convolutional layers and number of fully-connected layers, number of iterations to achieve 100% accuracy on training data, batch size and dropout layer. The results show that the proposed and modified AlexNet model with less number of layers has achieved the testing accuracy of 90.48% while the original AlexNet model achieved 86.65% over the untrained benchmark fish dataset. The inclusion of dropout layer has enhanced the overall performance of our proposed model. It contain less training images, less memory and it is also less computational complex.