Imageinterpretation depends on priori knowledge and contextual knowledge, actionsto perform to accomplish the specified goalsthe interactivity of the platform with its environment the cognitive vision platform has to beable to adapt its behavior by taking into account enduser specifications. That is why the development oflearning capabilities is crucial for cognitive vision systems.. All these interpretations are correct and enableus to conclude that semantics is not inside the image. When we look at the image on the left of figure 3, wehave to answer to the following question what is the semanticcontent of this image?

Imageinterpretation depends on priori knowledge and contextual knowledge, actionsto perform to accomplish the specified goalsthe interactivity of the platform with its environment the cognitive vision platform has to beable to adapt its behavior by taking into account enduser specifications. All these interpretations are correct and enableus to conclude that semantics is not inside the image. That is why the development oflearning capabilities is crucial for cognitive vision systems.. Our approach for the semantic image interpretation probleminvolves the following aspects of cognitive vision knowledgeacquisition and representation, reasoning, machine learning andprogram supervision.

Currently, we havethe design of minimal architecture more than solution for specificapplication,the platform is modular system which provides reusable and generic tools for applicationsinvolving semantic image interpretation probleminvolves the following aspects of cognitive vision knowledgeacquisition and representation, reasoning, machine learning andprogram supervision. We want to design generic and reusable cognitive vision platformdedicated to semantic image understanding. In particular, ahigh level language based on an ontology allows to describe new classes of objects. The work on ontological engineering presented above takes part on this requirement.