Sensor Anomaly Detection

EdgeSense-1 is the first model on this site that Edge Attention trained rather than converted: a small autoencoder that learns what normal looks like for a meter stream and flags what it cannot explain, from spikes and outages to a sensor stuck on one value. It is 12 kilobytes, and your data never leaves the browser.

EdgeSense-1 · 12 KB · trained by usRuns in your browserEnergy and equipment telemetryNo Edge Attention servers

or upload a CSV

CSV: one reading per line, last numeric column is used. Sample is a fictional 10-day meter stream.

EdgeSense-1 was trained by Edge Attention. Data is processed on your device and never uploaded.