Standalone mode performance

martin181818

hi community,

depthai-experiments/gen2-cumulative-object-counting at master · luxonis/depthai-experiments (github.com)

I am working on this project and it is working in normal mode and in standalone mode. The only difference here is that during the standalone mode when I get the results at the host and try to project : the video is more laggy and the fps value is slow. Is this normal and related to the encoding and sending the data to the host ? and is there a difference in the performance of the standalone mode because the video is delayed ( for example the number of detected objects ) ?

erik

Hi martin181818 ,
In standalone mode it's waiting for both available frame + tracklet before it does the calculation and sends the frame + metadata via TCP to the host (code here).
In host mode it won't wait for new tracklet - if there is a new frame it will just take the latest metadata and display the metadata on the frame. That's why you get higher FPS in host mode.

martin181818

erik hi Erik
Thank you for the reply. Okay I see. The fps is reduced due to waiting for both the frame and the metadata.
But, is the tracking working the same from host mode to standalone mode ? or is there a performance reduction ( not the same quality of detection and tracking ) during the standalone due to the lower fps ?

erik

martin181818 Tracking quality should be exactly the same, just visualization of the results could be slower. You could update the code a bit to not wait for both frame + tracklets in standlaone mode, so you would have faster video output.

martin181818

erik Thank you for the support.