An Empirical Study on Performance Evaluation in Automatic Image Annotation and Retrieval

Advances of information and communication technologies allow the creation of image archives extensively. As a result, the size ofimages database archives is increasing rapidly. So an efficient image annotation and retrieval system is highly desired. Automatically assigningkeywords to images allows one to index, retrieve and understand large collections of data. Many techniques have been proposed for imageannotation in the last decade that gives reasonable performance on standard dataset. However most of these works fail to compare their methodwith other methods that justify the need for more complex models. In this work, we compare the performance of various image annotationmethods, and propose that new base line method is that which outperforms the current state of art methods on two standard and one large webdata set