Three Feet

Three Feet is a camera-based assistant that monitors the space around a heater and warns when an identified object remains too close. With a compatible smart plug, it can also request power off. Objects such as blankets, clothing, paper, and cardboard can end up near heaters during everyday use. Three Feet helps users notice these changes using a supported camera feed, including a webcam or a repurposed phone. The application estimates floor geometry from the heater’s measured base dimensions and floor-contact corners, then displays an adjustable clearance zone. SAM 3.1 segments nearby objects, Muse Spark inventories the scene and assesses hazards, and Muse Wonder extracts clearance information from manuals for human review and summarizes incidents. Unclassified movement waits for model identification before starting an object-shutoff countdown. Recognized people and clothing being worn are excluded. Once an identified object remains inside the configured zone for its hazard-specific duration, the application sends an OFF request and records what triggered it. Local tracking follows objects between scene-triggered and periodic model checks to manage latency and API costs. The dashboard includes live overlays, approximate distances, browser notifications, voice alerts, plug status, and an event log. Three Feet is a prototype: distance accuracy depends on calibration, and a power-off request records a command rather than independently proving physical shutdown.