AI system facilitates people with blindness and low vision in interpreting and experiencing unfamiliar environments
摘要
Engaging with nature significantly enhances well-being, yet millions of individuals with blindness and low vision (BLV) are often excluded from these benefits due to constrained environmental perception. Here, we introduce VIPTour, an AI-driven system powered by the FocusFormer algorithm, which transforms complex scenes into structured, personalized graphs using tailored attention mechanisms and a BLV-in-the-Loop Adapter. Through intuitive, user-centered interaction, VIPTour facilitates active exploration and in-depth comprehension during dynamic sightseeing, while enabling accurate, long-lasting recollection and effective communication among BLV individuals post-journey. Extensive experiments demonstrate that VIPTour significantly enhances positive emotions and memory retention, with a 67.9% increase in positive emotional response, a 94.7% rise in arousal, a 772.73% improvement in cognitive mapping accuracy, and a 200% enhancement in long-term memory accuracy. These results underscore VIPTour’s ability to deliver an unparalleled, enjoyable, and memorable experience, promising profound benefits for the BLV community.