Edge and cloud computer vision suit different requirements. This page sets out where each applies.
Cloud deployment
Cloud CV is the right answer when your site has reliable high-bandwidth internet, when you need large models that do not fit on edge hardware, and when your events are not time-critical. If you are counting cars in a car park overnight, cloud is fine.
Edge deployment
Edge wins when response time matters (queue agents, PPE gates), when the network is unreliable (construction, regional, transport), when data minimisation is a compliance requirement (public space, workers), or when video egress cost would dominate the budget. Most Australian deployments we see fall into at least two of these categories.
Hybrid architecture
The most common architecture we end up deploying is hybrid: perception and real-time action on the edge, aggregation and long-term analytics in a Sydney-hosted control plane. That gives operators the latency and compliance of edge, with the analytics convenience of cloud — without the cross-border problem.
Three questions to ask your vendor
- Does raw video leave my site?
- How fast is the response time on a real-time agent?
- Where is the data hosted, and can you show me?
If your current vendor is streaming raw video to an offshore region, ask what it would take to stop.