Direcionamento automático da atenção em vídeos de realidade virtual

Abstract

The omnidirectional nature of 360º Virtual Reality (VR) videos offers users a highly immersive experience but also creates challenges related to attention and orientation, as important scene elements may be overlooked. This dissertation investigates automated attention guidance in immersive 360º VR videos through the integration of Natural Language Processing, Computer Vision, and adaptive visual effects. The research comprises three interconnected studies. The first explores the use of natural language video roadmaps, object detection, and segmentation for automated attention guidance. The second introduces Focus360, an architecture that combines roadmap interpretation, object detection, object tracking, and visual effects to improve guidance robustness. The third evaluates different object detection and tracking architectures using quantitative metrics and qualitative assessments. The results demonstrate the feasibility of converting natural language descriptions into visual attention cues and highlight the potential of combining computer vision techniques and adaptive visual effects to support user orientation in immersive environments. This dissertation contributes to the development of intelligent attention guidance systems for 360º VR experiences.