The first architectural decision in a video-analytics project is where video is processed: on an edge device near the cameras, on servers on-premise, or in the cloud.
Definitions
- Edge: analysis runs on a compact, accelerated device in the camera network (or in the camera itself).
- On-premise: analysis runs on servers in the organisation’s own server room or data centre. Together with edge, it forms the “video never leaves the site” approach.
- Cloud: video is uploaded over the internet and analysed in a cloud environment.
Comparison
| Criterion | Edge / on-premise | Cloud |
|---|---|---|
| Privacy | Video stays in your network | Video goes to third-party infrastructure |
| Bandwidth | Low — only events leave | High — continuous upload |
| Latency | Low — alerts generated on site | Depends on connectivity |
| Internet outage | Analysis continues | Analysis stops |
| Upfront cost | Hardware investment | Low |
| Scaling | Add hardware | Change service plan |
Which one fits?
- Plants, warehouses and banks with many cameras: edge / on-premise.
- Small, distributed sites with few cameras and lower privacy sensitivity: cloud can work.
- Multi-site organisations: process on-site, aggregate events centrally (hybrid).
Data protection
Keeping video on-site simplifies data-transfer and processor questions under GDPR or KVKK. Hype Vision recommends on-premise processing by default, with a hybrid option for central dashboards. See industrial computer vision for the wider picture.
Not sure what is feasible on your site? Send a few camera frames and your use case.
Evaluate your camerasFrequently asked questions
How many cameras can an edge device handle?
It depends on the device, frame rate and number of analytics per camera. Sizing is done during discovery.
How is an on-premise system supported remotely?
Typically through a customer-approved VPN or remote-access method, with access scope agreed in advance.