Génération d’explications textuelles en XAI : le cas de l’annotation sémantique

Semantic image annotation is an important field in which deep learning excels. However, an explanation of this annotation may be required in some application domains, like security or medicine. In the field of Explainable Artificial Intelligence, we study in particular an explanation which is a sentence in natural language that is dedicated to human users. It should provide them with clues about the process that leads to the decision : the labels assignment to image parts. We focus on semantic image annotation with fuzzy logic. It has proven to be useful when it comes to capture both image segmentation imprecision and the vagueness of human spatial knowledge. In this paper, we present an algorithm that generates textual explanation of the semantic annotation of image regions.