Comparative Analysis on Sediment Classification using Convolutional Neural Network

In this proposed work, we trained the neural network models for feature extraction and categorize the underwater seafloor sonar images into sediments, sands, mines, rocks, mud, gravels, etc using Convolutional Neural Network such as ResNet50, Inceptionv3 and VGG-19. Moreover, the performance of neural network is estimating by metrics measures such as validation accuracy, accuracy, loss, and accuracy loss upto 10 epochs. Hence assessment had performed among all three neural network models. The experimental output signifies that ResNet50 method achieves 94.4% appreciable accuracy with loss as 0.01 in distinguishing sonar imagery into sediments, sands, mud