Zabis Tech logoZabis Tech
Plan · No. 02

Stream Planner

Size an AI video analytics deployment on the NVIDIA DeepStream stack before you buy a single box.


Your cameras

Planning result

0 load units (1080p30 detection streams)

0 Mbps inbound

— TB storage

Deployment layout

Hardware classes compared

ClassCapacity (units)Units neededAvg. utilisation
Capacity figures are generic rule-of-thumb classes, not vendor benchmarks, with 70% target utilisation.
Engineer beside a compact edge computing box with a security camera in the background.
Analysis runs on an edge box beside the cameras.

Product notes

Our camera analytics product runs on the NVIDIA DeepStream SDK. Streams are decoded and analysed on an edge box next to the cameras. Only events and counts travel to the dashboard, which keeps uplink needs small.

How the estimate works. Each camera is converted to a load in units of one 1080p, 30 fps detection stream. Tracking multiplies that by 1.3 and attributes by 1.8. Bandwidth assumes about 4 Mbps for 1080p30 H.264, scaled by pixel rate, and 40% less with H.265.

Limits. Real performance depends on model, hardware, scene and tuning. Treat these numbers as planning estimates and validate with a pilot on your own streams.

Request a demo or a licence

The camera analytics product runs on NVIDIA DeepStream and can be licensed or deployed for your site. Send your planner numbers and we will arrange a demo or a pilot.