Block 3. Edge + Cloud AI

1

Edge vs Cloud Analytics: The Deployment Decision That Defines the System

Edge, edge-server, or cloud? Compare where video analytics software runs on latency, bandwidth, cost, accuracy, and privacy, with worked 2026 math.

2

On-Camera (Edge) AI: What Runs on the Camera and Why

What an edge AI camera actually runs: the NPU inside, the small models that fit, the metadata it emits over ONVIF, and the limits of on-camera compute.

3

Edge Servers and On-Prem AI Appliances

What an edge AI server runs: GPU stream density, when an on-prem appliance beats camera and cloud, and the latency and data-residency it buys you.

4

Cloud Video Analytics: When to Send Video Up, and What It Costs

What cloud video analytics really costs: the per-minute API trap, GPU rental, storage and egress meters, and when sending video to the cloud pays off.

5

The Hybrid Processing Pattern: Split Work Between Edge and Cloud

The hybrid edge-cloud pattern: run detection and recording locally, send the cloud only events and clips. What runs where, the cascade, and the seams.

6

The Economics of Analytics: Bandwidth, Compute, and Storage Tradeoffs

Price video analytics per camera: the bandwidth, compute, and storage levers across edge, server, and cloud — plus the break-even points that decide it.

7

Latency and Accuracy at Each Tier

Detection latency and accuracy by surveillance tier: edge answers in tens of ms, cloud in hundreds, and accuracy is a range — when milliseconds matter.

8

The Edge + Cloud Reference Architecture

The complete edge + cloud reference architecture for AI video surveillance: all six layers, the ONVIF standards spine, and the bandwidth and cost budget.