Record SVO w/o CUDA w/o depth


We are currently trying to find out the minimal hardware requirements for our use case - to record SVO files including only the left+right camera streams from a ZED 2 and send it over network to a more powerful edge computer that computes the depth map.

We know that we can use a Jetson Nano for this task, but we are also wondering if we could use an even more resource constrained device without Nvidia GPU for this task? Reducing price and power consumption is the key thought here.

On this page is stated that no Nvidia GPU is required, but I can not compile the “svo recording” sample without CUDA installed on a Raspberry Pi 4 running Ubuntu 20.04.

I also found the Open Capture Camera API, however, it does not support recording an SVO file, which we need for the edge computer to open it with ZED SDK.

Can anyone please point me in the right direction for this?

The only way to record a SVO without using the ZED SDK is by using the tool ‘ZED Explorer’ with command line, see:
usr/local/zed/tools/ZED_Explorer -h
You should be able to install the ZED SDK on your raspberry to use it (Cuda will not be installed and the API won’t work but the Explorer will).

This works, thank you!

Do you also know if there is a way to enable hardware encoding without Nvidia GPU?
Raspberry Pi 4 should support H.264, but when I set this compression in ZED_Explorer I get:
Cannot record SVO file (unsupported compression). Make sure your hardware can support the choosen compression.

I guess it tries to use some Nvidia library for that?

Indeed the current encoder of the Explorer is using the Cuda API so you can’t use it.

Ok I see, anyway, thanks for that info!