Depth information without SDK

jonasf · February 17, 2021, 2:23pm

Hi,
I’m looking for a depth camera for robotics application. ZED2 camera fulfill every of my needs except potentially one.

Our application needs to run on an embedded target such as a Raspberry Pi board and I was wondering if ZED2 camera could provide depth information without using the SDK (I saw that ZED SDK was not available without CUDA dependencies).
There is a “zed-open-capture” repo on github indicating a library that doesn’t require CUDA, but I can’t figure out if it provide depth information too.

Thanks for answering me.

Myzhar · February 17, 2021, 2:54pm

Hi @jonasf
the zed-open-capture repo can be used to extract depth information using for example the OpenCV library.
There is an example explaining how to rectify the images, it can be enhanced by adding depth estimation.

The only problem is that an embedded board without CUDA capabilities could be not enough for a robotic application, the ZED publishes images with high resolution and high framerate and the Stereo Vision algorithm is highly demanding in terms of elaboration costs.
This is why the ZED SDK requires CUDA.