Analyse de l'application EchoSense-AI pour le 8 octobre

EchoSense-AI is an open source, accessibility-focused visual assistance app. It combines the camera, local computer vision, speech, OCR, and an on-device multimodal language model to describe surroundings and make printed information easier to access. Features • Assistant provides multi-turn text or voice chat and can analyze a camera photo, stored image, PDF, or text document. • Navigate describes the current scene and accepts optional voice commands such as “help me find my keys.” • Safety combines object detection with ARCore metric depth and vertical-plane checks when supported, then falls back to camera-relative proximity estimates. • Currency identifies visible bank notes. • Read extracts document text from the camera or an imported file, reads it aloud, and offers Guided Scan with spoken page-edge guidance, automatic stable-page capture, perspective correction, image enhancement, and a manual capture fallback. It also identifies the approximate center color, reports camera-relative light level, and scans common barcodes and QR codes entirely on-device. Its accessible Magnifier provides zoom, freeze, flashlight, contrast, inversion, and grayscale controls. • Translate extracts document text and translates it to English. • On-device analysis uses a downloaded Gemma multimodal model without sending camera images to an AI provider. • Optional online analysis supports user-provided Gemini or OpenAI API keys and configurable ask, fallback, or prefer-online behavior. afety warnings are experimental estimates based on camera detections. EchoSense is not a certified mobility aid and should not replace a cane, guide dog, attentive travel, or other established safety practices.