Running two model in one time

How well does Jetson optimize with these two together?

  1. VideoMAE. For video-classification, fine-tuned from videomae-base-finetuning-kinetics;
  2. YoloV5 or YoloV8. With person detection and in the future with more different detections.


Do the two pipelines use the same input source or not?
If yes, please check our Deepstream SDK.
If not, you can just create two pipelines and let the GPU scheduler accommodate the task.


Overall yes. With one input source. But for VideoMAE whole image will be cropped for three-four pictures with different xyxy, so they will show different places on the source, and after they will come into the model.


Please check if Deepstream can support your use case.

Yes, of course, I will work on it next week.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.