The semantic orientation in space system is a concept of combining spatial data processing technology and semantic analysis. Its main purpose is not only orientation in space but also understanding the meaning of objects in a given environment. Semantic orientation in space could be used, for example, to create more advanced navigation systems that understand the context of natural language (e.g. “drive up to the blue chair on the right”, “how many whiteboards are there in the rooms”). The realization of the goal should be a multi-stage process combining the aspect of mapping with semantic analysis and aspects of navigation (implemented by a separate team), ultimately creating a complementary whole. The project concerns stage 1, which will allow for the verification of the basic concept of semantic analysis of space and the introduction of semantic representation of objects. This stage consists of the following tasks:
- developing assumptions and application scenarios, as well as a verification method
- developing a method for generating accurate and reliable linguistic descriptions based on visual data
- developing a method for semantically embedding and searching text data, and linking it to image data
- development of a method for analyzing user queries regarding the search for target navigation points based on natural language queries (which will enable navigation to a specific point based on semantic features)
- development of an API that will enable image processing (e.g. from navigation and mapping subsystems) and searching and locating objects on a map, which will allow for a wide application of the developed technologies in various navigation scenarios, as well as verification of the scenarios adopted in the project.