I found nothing about how DepthAI using Myriad X VPU to generate depth images on Luxonis's github, is it open-sourced?
And is there any resolution or color-format (mono/colored) limitation about DepthAI?
For example, is it possible to generate depth image from our customized stereo cameras with 4k resolution using DepthAI?

    Hi Lala ,

    I am checking with the team and will circle back with you. Thanks!

      Hi Lala,

      Unfortunately, we cannot release our firmware for the Myriad X under an open source license.

      Can you give a little more detail on your custom cameras? Are they based on a common image sensor?

      Best,
      Philip

          Hi philipmolloy ,

          Thanks for your response.
          We haven't decided what camera module to use yet.
          We know that the OpenCV AI Kit OAK-D uses depthAI solution and provides depth streaming at 720p, 120 fps.
          So we want to know whether is the DepthAI SoM meet our requirement : higher resolution ( 1080p or 4k ) with lower fps.

              Lala

              So the depth engine in depthai/OAK is limited to 1280 resolution width. It technical would be possible to decimate larger images into this resolution, and produce a depth map from these higher res sensors across the whole field of view, but it would technically still be 1280 actual horizontal resolution (instead of 1920 or 3840 horizontal pixels).

              That said, it is likely possible to do such a decimation (to produce the entire image) and then fill this in with a tiled full resolution. That said, we have not implemented such a thing. It is likely possible with our Gen2 pipeline system, with the host involved for helping to shuffle tiles/etc. back/forth.

              So TLDR, likely doable - but we haven't tried it. And definitely not just plug/play doable. I suspect 1080p could run at 30FPS. And I'd expect 4k would be something like 10FPS or so. But I don't know how much host load this would cause.

              Thoughts?

              Thanks,
              Brandon

                Understood, thanks for your detailed explanation.

                Best,
                Lala