Better object recognition using bag of visual word model with compact vocabulary

Searching the information from the web has attracted the attention of industries and researchers, relaying mainly on text based queries. Text based queries are foundation of visual or text based information for retrieval. However, query as visual content also getting popularity. The main problem for visual query is the computational cost for searching. A number of text based inspiring frameworks were proposed for said objective. Bag of word model is one of the prominent and state-of-the art method for text retrieval, also widely used in visual search known as bag of visual word. In this paper, we proposed bag of visual word model based framework for image retrieval that encodes the geometric information of given point in the image. Our bag of visual word representation is relatively compact and effective. Experiments on benchmark dataset, Oxford 5-K, show the effectiveness of proposed framework.