An efficient image recognition framework for content based image retrieval using deep learning techniques

Abstract

In recent years, trillions of images have been shared on various types of social media platforms on an everyday basis. Social media s wide range of features dominates the globe by making the searching process complex; identification of relevant objects or images turns out to be impossible due to the varying redundancy levels. High-level image visuals are preferably represented in the form of feature type vectors comprising of numerical values, and hence lack semantic representation of image features. Content Based Image Retrieval (CBIR) can differentiate the images based on lower level factors like shape, spatial layout, color, and texture. The semantic gap is reduced by using machine learning techniques in existing studies. The CBIR approaches are available, which use the classic similarity measure that focuses on extraction results but less on computation time and computational complexity. Thus, the CBIR system needs effective and efficient image retrieval with minimum time and computational complexity. Deep learning techniques are growing wider day by day in the process of Content Based Image Retrieval (CBIR). The recognition of the image is based on its shape, attributes, and tag. It is challenging to establish the connection between semantic ideas in the vast real-world applications. Social media is dominating the globe with its wide range of features, and people are finding difficulty in choosing suitable objects or images because of any redundancy. So the proposed method is based on content-based image recognition and tagging using deep learning techniques. The tagging of the image is used here for easy identification of the objects. The Geon similarity model is used to extract the maximum similarity of the different images with its accurate and rapid computation methods. newline