Content-based Image Retrieval based on Convolutional Neural Networks

Images can communicate a service, brand or product. Moreover, images provide depth and context to a description or story and give a much more intense experience than writing alone. Image retrieval is the highest searching performance process, especially in large databases. Search by content is a powerful tool for many industries. Images are more complicated than texts. Each image has important features like color, texture, and edges. Searching by part of image is termed content-based image retrieval. Convolution neural network achieved high accuracies in images feature extraction and classification. For high-performance image retrieval, we proposed an intelligent model using CNN. The model was applied on Cifar10 and Mnist datasets. The proposed model achieved 92.9% and 99.8% accuracies for Cifar10 and Mnist datasets respectively. The model achieved the highest accuracies and best processing performance compared by State-of-arts-models.