TIA-INAOE's approach for the 2013 Retrieving Diverse Social Images task ∗
Hugo Jair Escalante, Alicia Morales-Reyes · 2013
This paper describes the approach adopted by the TIA-INAOE team for the 2013 Retrieving Diverse Social Images task of MediaEval. The challenge consists in re-ranking a list of images returned by a retrieval system in such a way that visual diversity among images in the first positions is maximized [5]. A database of partially-annotated images of tourist destinations is provided. The problem is tackled as an optimization one with the aim of finding a new image ranking that shows improved diversity according to the criterion defined by MediaEval. The proposal is to apply a multi-objective evolutionary algorithm to simultaneously maximize image diversity in consecutive positions of the list and minimize divergence from the original list. Multi-modal information can be incorporated by the proposed approach. Results obtained in the MediaEval forum are reported and analyzed.