System and method for on-device person identification and provisioning of smart alarms

This document describes systems and methods for on-device person identification and provisioning of smart alarms. The system includes a decentralized multi-camera system for face recognition on a device. A device (e.g., a security camera, video doorbell) captures an image/video of a person, processes input image frames, detects facial images, filters static faces, and aligns the rotation of the faces to upright and frontal. The device then filters low quality face images and/or images with most of the face being occluded. The device computes a facial insert, compares it to a set of locally stored reference inserts, and sends a matching result to a cloud service, which notifies the device owner of whether the observed person is a known person or a stranger based on the matching result. Face detection and recognition calculations are performed on the device, rather than on the cloud. No sensitive information is transmitted from the device, and thus privacy is protected. 本文档描述了用于设备上人员辨识和智能警报的供应的系统和方法。该系统包括用于设备上面