Mi SciELO
Servicios Personalizados
Articulo
Indicadores
- Citado por SciELO
Links relacionados
- Similares en SciELO
Compartir
Revista Cubana de Ciencias Informáticas
versión On-line ISSN 2227-1899
Resumen
SANCHEZ SANTIESTEBAN, Sergio. Content-based image retrieval using descriptors generated by Convolutional Neural Networks. Rev cuba cienc informat [online]. 2018, vol.12, n.4, pp. 78-90. ISSN 2227-1899.
Content-Based Image Retrieval systems allow to search and retrieve images that are similar to a given query image using features for representing the visual content of the images. In this work it was developed a method to retrieve digital images indexed in databases using its visual content, without textual annotations. Automatic descriptions of the contents were obtained using deep neural networks. Pre-trained Convolutional Neural Net- works were proposed to create global descriptors. Dimensionality reduction techniques were applied to increase the efficiency in performance. Results obtained by this method, over two publicly available datasets, were better than performance of traditional methods and comparable to other approaches based on deep learning which are the state of the art in Content-Based Image Retrieval. Proposed method could be extended by the addition of stages of feature integration with a greater degree of abstraction.
Palabras clave : Convolutional Neural Networks; global descriptors; image retrieval; information retrieval.