EchoSight: AI-Powered Assistive Vision Agent

EchoSight is an AI-powered assistive vision platform designed to help blind and visually impaired individuals understand and interact with their surroundings through computer vision, spatial intelligence, and natural voice interaction. The Problem Navigating unfamiliar environments presents significant challenges for people with visual impairments. Traditional mobility aids provide essential physical feedback but cannot describe complex scenes, identify distant objects, or answer contextual questions about the environment. Our Solution EchoSight combines local computer vision, AMD-powered AI models, Evolus intelligent agents, and a conversational voice agent to transform visual information into accessible spoken guidance. How It Works EchoSight uses YOLO to detect objects and Depth Anything V2 to estimate depth. A spatial reasoning engine analyzes object positions and potential obstacles, generating navigation warnings locally without depending on cloud connectivity. For deeper environmental understanding, AMD Token Factory provides multimodal AI inference, while the Evolus Accessibility Assistant interprets grounded visual observations to generate contextual responses. A conversational voice agent allows users to ask natural questions such as “What is in front of me?” or “What is behind those people?” and receive spoken answers. Intelligent Architecture EchoSight uses a hybrid edge-cloud architecture. Time-sensitive obstacle warnings are processed locally, while cloud-powered AI handles more complex scene interpretation. Session isolation, observation freshness checks, and cloud-failure fallbacks improve reliability. Our Vision EchoSight demonstrates how computer vision, intelligent agents, and accessible voice interfaces can work together to make environmental information more accessible and support greater independence.