One of my client need to perform video analytics on 100 RTSP camera.
Which GPU is recommended to stream and run inference on 100 RTSP cameras in Deepstream.
And what is the network infrastructure required.
And is single deepstream app capable of streaming and infer on 100 cameras or do we have to create new instance of same deepstream app each with some camera streaming and inference.
The Data center GPU - GA100 should meet this requirement.
Please check Performance — DeepStream 5.1 Release documentation (nvidia.com)
Thanks for the link
Let’s say I take T4 GPU with the system suggested, and set cameras at 30fps and stream type h265, according to document I can stream 64 cameras which utilises 58% GPU.
Couple of questions,
1-Will all 64 cameras run in single Deepstream app and I can access metadata for all in probe function ?
2-if GPU utilisation is 58% can I add more cameras to same or not because of memory constrain won’t be able to add more ?
3-Is it possible to get the network block diagram or any networking details through which these benchmarks were performed to try at our end i.e. to stream 64 cameras ?
Reiterating 64 camera can run, so were they running in Deepstream ?
And regarding network was as to how much bandwidth what type of routers specs, POE switches, but will test and get an idea
can you share where you get these info?
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.