“Spatial Sense” is a navigational aid project for individuals with low vision, utilizing the synergy of Multimodal Large Language Models (LLMs) with “Segment Anything,” “Depth Anything,” and “Recognize Anything” models for real-time environmental object detection and obstacle negotiation. Central to the project is an LLM agent, engineered to operate on edge devices like Qualcomm Snapdragon boards, ensuring quick, on-device processing for immediate user feedback.
