Deep Features Selections with Binary Marine Predators Algorithm for Effective Classification of Image Datasets

N. Muhammed Noori, Omar Saber Qasim ^*

Department of Mathematics, University of Mosul, Mosul, Iraq

Emails: NoorMuhammedNoori@uomosul.edu.iq; omar.saber@uomosul.edu.iq

Abstract

The paper proposes a method for improving the accuracy of image classification by combining CNNs and the Binary Marine Predators Algorithm (BMPA). The CNNs used in the study, ResNet 50 and AlexNet, were trained on ImageNet and used to extract features from the images in the dataset. Features are taken from layers (avg_pool) in ResNet 50 and (drop7) in AlexNet. These features were then fed into the BMPA algorithm, which selected the most relevant features and removed irrelevant ones to improve the classification process. The proposed method is said to be efficient, capable of achieving higher classification accuracy, and able to select the best features. The authors believe that this approach could be applied to a variety of other image classification tasks. It is important to note that the effectiveness of this method should be evaluated on a range of datasets and compared to other state-of-the-art methods.

Keywords: Convolution Neural Networks; Marine predators algorithm; Feature selection; Classification.