ImageNet Classification using ResNet50 and Binary Marine Predators Algorithm

• Published in: 2023 International Conference on Information Technology, Applied Mathematics and Statistics (ICITAMS) pp. 340 - 344
• Main Authors: Noori, Noor Muhammed, Qasim, Omar S.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 20-03-2023
• Subjects: classification; Classification algorithms; convolution neural network; Convolutional neural networks; Feature extraction; feature selection; Filtering algorithms; marine predator algorithm; Metaheuristics; Support vector machines; Vectors
• DOI: 10.1109/ICITAMS57610.2023.10525313

Image classification accuracy relies heavily on the quality of features used to classify the dataset. Researchers commonly use metaheuristic algorithms like the Marine Predators Algorithm (MPA) for feature selection and dimensionality reduction to enhance the accuracy and reduce the number of features required. In this study, we used a ResNet-50 convolutional neural network (CNN) pre-trained on ImageNet to extract features from the image dataset. We then utilized the Binary Marine Predator Algorithm (BMPA) to filter out irrelevant features and select the most pertinent ones from the feature vector extracted from the "avg_pool" layer in ResNet-50. Our proposed efficient approach can achieve superior classification accuracy and optimal feature selection. We anticipate that the combination of CNN and BMPA can be utilized in various other image classification tasks to enhance their performance.