I want to analyze an image by several neural networks in an embedded system. My question is the following :
can we run in parallel, on graphic cards like Jetson TX2 or Xavier, several inference tasks. Each task using a different neural network. For example : A task using a YOLOv5 and an other task using a Mask RCNN.
Thanking you in advance for your response
The answer is yes, but there will be some possible performance impact.
In your case since your target platform is Jetson, I am moving this discussion to an area where there are a number of community members which would have similar requirements to you.
Suggest to use DeepStream SDK, please find more information from NVIDIA DeepStream SDK | NVIDIA Developer
Thank you for this answer.
One last question: If you use different networks, do you have to load and unload them from the memory before and after each use ?
Yes, DeepStream SDK is based on gstreamer and in each run, it loads and unloads the model.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.