MLCamera performance

[edgar.santos]

Hi everyone,

I’m having some questions about capturing performance. I have it working just fine, but higher resolutions degrade performance a lot.

For testing purposes, I’m using 3840x2160 at 15fps capturing using RGBA.
If I don’t do any processing in the capturing handler, the performance is stable.
However, if I show it on the screen, I need to upload the rgb data to a texture every frame. That is tanking performance. I’m already caching the texture object itself.

I’m wondering if there’s a way (any way) to capture higher resolutions without tanking the performance.

Thank you for your time

Unity Editor version : Unity2022.3LTS
ML2 OS version : 1.12
Unity SDK version : 2.6.0
Host OS : Windows

[kbabilinski]

Here is a post that could help:

Unity : Rendering YUV 4k Camera Stream

Here is a managed version of the visualizer that does not require enabling the unsafe flag in the project settings. Note the scripts are named the same so only one can be in your project at a time unless you rename the script and it’s reference in the AsyncCameraCapture script.

I recommend using YUV instead of RGB since using YUV reduces the overhead. If possible you may want to do the conversion on the receiving side.

[edgar.santos]

It looks very promising, thank you very much!