Ronghang Hu, Marcus Rohrbach, Jacob Andreas, Trevor Darrell, Kate Saenko, 2016, Modeling Relationships in Referential Expressions with Compositional Modular Networks, Web, url: https://arxiv.org/abs/1611.09978
Replication of the above paper as a machine learning class ter project
Localization of referential expressions in an image
Generally, when we are talking about some people or some object, we refer them accordingly their relationship with the other objects or people. However, if we, as a human, understand objects or people with these relationships, then computers also must be able to understand this way so that we can have more accurate results. In this paper, I use Computational Modular Networks (CMNs) which first parses the referential expression as entities and relations,then identifies in the image and finds the most related one.