Convolution Neural Networks for Content based Image Retrieval Using Quantum Cuckoo Search Optimization Technique

Abstract

The goal of Content Based Image Retrieval is to get images from a vast collection that is related to a query image or relevant text format based on their visual content. An efficient retraining approach for improving Convolutional Neural Network (CNN) features for Content Based Image Retrieval (CBIR) is proposed in this study. CBIR is one of the fields of computer vision that is quickly growing. Convolutional neural networks (ConvNets or CNNs) are one of the primary types in neural networks. CNNs are frequently used in identification of images, classification of images, detection of object, recognition of face, and other fields.To achieve the best results, go one step further and handle the CBIR problem with CNNs, which are capable of not only categorising but also accurately localising pictures of distinct classes based on the requirements. Optimization methods are used to change the internal parameters of processing layers. In our research, the quantum cuckoo search optimization (QCSO) approach is used to identify pictures and produce results. The findings for CNN-based image retrieval with 94.8 percentofoverall accuracy and QCSO-based image retrieval with 96.4 percent of overall accuracy are displayed in comparison. Matlab is used to generate the experimental findings.