Content Based Multimodal Retrieval for Databases of Indian Monuments

With the explosion of multimedia content online image search has become a viable way of retrieving relevant images. However, current methods of image search use textual cues to retrieve images and do not take into account the visual information they contain. In this paper we aim to crawl and build multimodal image search engines that take into account both the textual and visual content relevant to the images. We intend to use off the shelf text search engines to accomplish the above task, making the construction of image retrieval systems extremely easy. We build visual models by using the bag of words paradigm and propose and validate through experimentation a combined multiple vocabulary scheme that outperforms normal vocabularies.