A Comprehensive Survey of Deep Learning Algorithms in Coral Reef Segmentation

Main Article Content

A. Shakul Hamid, M. H. Ibrahim, J. Jannathul Firthous

Abstract

Image segmentation carries a significant contribution in scene understanding.The most recent semantic segmentation technique can automatically detect the coral reefs and are addressed upon using deep architectures like Convolutional Neural Network, Encoder-Decoder architecture, Recurrent technique gains much attention among researchers in computer vision implementations especially for the Neural network etc. Semantic segmentation takes each input image and assigns a class ID for every image pixel. It involves monitoring and preservation of marine ecosystem. The segmentation and classification of underwater creatures is clustering the similar parts of images. For instance, the left side of Fig 1 illustrates the original image, and on the right , is the challenging to trade-off between the accuracy and semantic segmentation of the image[2]. Computational efficiency. Deep   Learning(DL) algorithms outperforms the traditional techniques in increasing the performance. This paper reviews the promising deep learning architectures for segmenting seawater creature like coral reef using semantic techniques. Firstly, it explains about the background concepts and then the existing methods like Fully Convolutional Network(FCN), SegNet, DeepLab, ReSeg along with the highlighting contributions are presented. Finally, the survey of segmentation models in coral reef images is summarized.

Article Details

Section
Articles