Intelligent Content Based Image Retrieval Model Using Adadelta Optimized Residual Network

Content based image retrieval (CBIR) system produces useful representation of images with the consideration of the visual features of the images like color, texture, shapes, etc. An important requirement of CBIR is ensuring useful retrieval of images for the applied query images (QI). Therefore, this paper presents a new CBIR model using Adadelta optimized residual network, image retrieval against query images. The proposed model involves ResNet 50 based feature extractor to derive a useful set of features. Besides, Adadelta optimizer is applied to effectually tune the hyperparameter of the ResNet-50 model to improve the retrieval performance. In addition, Euclidean distance is employed as a similarity metric to identify the highly similar images that exist in the database with respect to the applied QI. The use of Adadelta optimizer helps to considerably boost the retrieval outcomes. For validating the improved retrieval outcomes of the proposed model, an extensive set of simulations take place on benchmark datasets and the obtained outcomes highlighted the supremacy of the proposed model over the other techniques.