Abstract
Deep learning has become one of the top performing methods for many computer vision tasks such as images retrieval. It has been deployed so far to bring improvements to learning feature representations and similarity measures.
In this article, we present a new search method to represent and to retrieve images based on the vector space method, called vectorization. This method transforms any matching model of images to a vector space model providing a score using the Convolutional Neural Networks (CNN). The results obtained by this model are illustrated through some experiments and compared with several state-of-art methods.