Skip to content

Web Interface

Every OpenIPC camera serves this interface itself, over its own address — there is no cloud account and no app in between. These are the pages of a camera running a current build: the live picture, recordings played back from the memory card, detection drawn over the picture, and the settings behind them.

Dashboard page of the OpenIPC web interface
Dashboard
Live page of the OpenIPC web interface
Live
Recordings page of the OpenIPC web interface
Recordings
Camera - Settings - Motion detection page of the OpenIPC web interface
Camera - Settings - Motion detection
Camera - Settings page of the OpenIPC web interface
Camera - Settings
Camera - Stream URLs page of the OpenIPC web interface
Camera - Stream URLs
Camera - Config file page of the OpenIPC web interface
Camera - Config file
System - SD Card page of the OpenIPC web interface
System - SD Card
System - Logs page of the OpenIPC web interface
System - Logs
System - Network page of the OpenIPC web interface
System - Network
System - Time page of the OpenIPC web interface
System - Time
System - Update page of the OpenIPC web interface
System - Update
System - Console page of the OpenIPC web interface
System - Console
System - File Manager page of the OpenIPC web interface
System - File Manager
Services - Telegram page of the OpenIPC web interface
Services - Telegram
Services - OpenWall page of the OpenIPC web interface
Services - OpenWall

The beach on the Dashboard, Live, Recordings and Camera - Settings shots is the view from a camera belonging to @usa-, an OpenIPC user, published with permission. The Motion detection shot is staged: the street is a frame of the recording in the video below, with faces pixelated, and its three boxes were handed to the interface's own overlay, because the lab camera's build carries no person detector.

Person detection and tracking

White flash: a detection arriving, about 2.5 a second. Coloured box and number: the same person, carried between detections. Faces pixelated.

Object detection (objectDetect in the settings) reports people in the picture. Every box goes where motion already goes: drawn over the picture in the browser, into the ONVIF metadata track and the track inside a recording, to pan/tilt following, and out of /api/v1/analytics.

On the GK7205V500, V510 and V530 it is YOLOv8 on the neural accelerator, with the models in the firmware, and it tells people from vehicles and animals (more).

Detections arrive a few times a second, not with every frame, and the video shows what makes that enough: a tracker that keeps each person's number through the gaps. Here it runs on detections made from the recording at that rate; it is a prototype, not yet in the firmware.