Glossary
What is edge AI video analytics?
Definition
Edge AI video analytics runs computer-vision models close to the cameras, inside the camera or on an on-site server or appliance, rather than sending video to a remote cloud for analysis. Processing on site reduces upload bandwidth, keeps video inside the local network and can continue working when the internet link is slow or interrupted.
Edge AI video analytics explained
Edge processing can mean in-camera analytics, which are limited by the camera's own processor, or an on-site server with GPU acceleration that analyzes many camera streams at once.
Hybrid designs analyze video at the edge and send only events and metadata to a central system for multi-site oversight.
OrionAlerts
How OrionAlerts handles it
In the standard OrionAlerts deployment, a Nebula OS node sits on site next to the cameras. Video and AI inference stay inside the site network, with GPU acceleration sized per deployment to the channel count and detections running, and only health, alert and configuration metadata sync to Orion Cloud. The on-site deployment is built to work through constrained or intermittent WAN links.
Explore AI Vision & Real-Time AlertsFAQ
Edge AI video analytics: common questions
What is the difference between edge and cloud video analytics?
Edge analytics processes video on site, inside the camera or on a local server. Cloud analytics uploads video, clips or frames to a remote service for processing. Edge reduces upload bandwidth and keeps video local; cloud reduces on-site hardware. The on-premise vs cloud comparison covers the trade-offs in more detail.
Is edge AI the same as AI cameras?
Not exactly. AI cameras run analytics on their own processor, which is one form of edge AI. Edge AI also includes on-site servers or appliances that analyze streams from many ordinary IP cameras at once, usually with GPU acceleration. OrionAlerts takes the second approach, running analytics on a GPU-accelerated node at the site.
Does edge AI keep working if the internet goes down?
Detection that runs on site does not depend on the internet link, so cameras keep being analyzed during an outage. Functions that need the link, such as remote viewing or notifications sent to phones and email, depend on connectivity. In OrionAlerts, recording and AI detection run on the on-site node.
Related terms
Video analytics
Software that analyzes video automatically to detect people, vehicles, objects and events, and raises alerts without constant human watching.
Read moreVideo surveillance as a service (VSaaS)
A cloud delivery model where recording, viewing and management of surveillance video are provided as a hosted subscription service.
Read moreRTSP (Real Time Streaming Protocol)
A network protocol used to set up and control live video streams from IP cameras and recorders.
Read more
See it on your own cameras
Book a walkthrough of OrionAlerts on Nebula OS, from camera connection to alerts on the dashboard, mobile app and email.