L’INTERPRETATION SEMANTIQUE DES EXTRACTIONS DU CONTENU D’UNE IMAGE

Abstract

Nowadays, three challenges of relationship-detection should be considered in order to build a strong model namely; long-tail problem, large intra-class divergence, and the semantic dependency or semantic gap. The aim of content-based image retrieval systems must provide maximum support in bridging the semantic gap between the simplicity of available visual features and the richness of the user semantics. Another issue is the long-tail problem where there is a low infrequent appearance of some objects (i.e. predicates) versus to the high occurrence of others. For that, an adequate scaling is demanded. The third problem that must be solved to build a strong CBIR system is the Intra/inter-class divergence. For the Intra-class divergence, objects are belonging to the same class but can’t be represented with the same visual characteristics, whereas the inter-class divergence is where the similar visual descriptors can relate to two objects that are not related to each other.

Nowadays, three challenges of relationship-detection should be considered in order to build a strong model namely; long-tail problem, large intra-class divergence, and the semantic dependency or semantic gap. The aim of content-based image retrieval systems must provide maximum support in bridging the semantic gap between the simplicity of available visual features and the richness of the user semantics. Another issue is the long-tail problem where there is a low infrequent appearance of some objects (i.e. predicates) versus to the high occurrence of others. For that, an adequate scaling is demanded. The third problem that must be solved to build a strong CBIR system is the Intra/inter-class divergence. For the Intra-class divergence, objects are belonging to the same class but can’t be represented with the same visual characteristics, whereas the inter-class divergence is where the similar visual descriptors can relate to two objects that are not related to each other.