Combining Deep Learning and Ontology to Reveal Video Sequences Semantics
Jemai Bornia, Frihida Ali · Revue d intelligence artificielle · 2021
As a semantic technology, an ontology is a formal hierarchical taxonomy of lexical terms including their syntactic and semantic relationships employed to represent the meaning conveyed by a knowledge domain.The huge increase in video data volumes creates a growing demand for efficient methods to reveal and manage their semantic content through an ontology.Therefore, the need for a system that allows efficient representation of video content is obvious.This paper proposes a method to extract ontology from video streams.It is based on Deep Learning technology to identify objects and movements in video scenes.All extracted features are stored in an OWL file.The OWL file will be used afterward to exploit video documents in many ways such as modeling, indexing, querying, and feeding ontology editors to visualize, elucidate and reason on the semantic structure.