QWO-IRV2 Deep Learning Model for Performing Retrieval of Multi Object Images

In the present scenario, data storage in the form of images is developing over the internet and effective techniques need to be implemented for retrieval of images. A large amount of data is available for numerous applications. Content-Based Image Retrieval (CBIR) is one of the important aspects that need to be considered. The goal of the paper is to retrieve multi object images from the database, which is a difficult issue for images that contain multiple objects. To overcome the problem, a deep learning technique along with an optimization method is developed for effective retrieval of multi object images. The Deep Convolutional Neural Network (DCNN) is designed. The network designed is ResNet 101 and Inception ResNetV2 (IRV2) for extraction of different types of features for the given input data. The features extracted depend on colour, shape, texture and so on. All the features are optimized using Quantum Whale Optimization (QWO). The final step is to estimate the similarity distance between the input query image and the data of images stored. Euclidean distance is utilized to identify the similarity index or level of similarity. The experimental evaluation is performed on the Corel database. The effectiveness and efficiency of the work are portrayed by utilizing metrics like recall, specificity, precision, and accuracy. The overall accuracy obtained by the proposed model is 98.33%. The suggested work QWO-IRV2 enhances image retrieval performance on the benchmark dataset in terms of accuracy, precision, and recall.