Automated Attention Guidance in Virtual Reality Videos

Download PDF
Open Webpage

Paulo Vitor Santana Silva, Lucas L. Neves, Rafael A. Goiás, Diogo Fernandes Costa Silva, Rafael Teixeira Sousa, Arlindo Rodrigues Galvão Filho

Published: 2025, Last Modified: 28 May 2026SVR 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Despite their immersive nature, 360° virtual reality (VR) videos often lack effective attention guidance, leading to user disorientation and missed information. This work proposes a novel method integrating computational vision with natural language processing to automatically guide user attention in 360° VR. It leverages natural language roadmaps to identify and track key elements, applying dynamic visual effects. The comparative evaluation identified Grounding DINO as a particularly suitable detector, while DAM4SAM and Segment Anything 2 (SAM 2) demonstrated strong performance for tracking. Demonstrated on a 360° VR tour, this approach can significantly enhance user experience and comprehension, advancing automated attention guidance for immersive content.