Date of Publication :25th April 2018

Abstract: Deep learning has given very promising results in the field of image retrieval. These results are generally built on the already existing deep learning systems that are used for the standard classification problem. A popular techniques in this domain employees binary hash codes which are used as the semantic representation of the image. The approach has a major bottleneck relating to the time of computation. Our idea is built around combining this approach and performing clustering based upon the hash code. The relevant clusters are then obtained by both direct matching as well as appropriate clusters within some small hamming distance.

Reference :