I would like to run inference on a batch of N temporally contiguous frames from a single source, but it seems that nvstreammux is designed to only batch across multiple sources at the current clock time.
Does nvstreammux support this type of batching? If not, can this type of batching be achieved without writing a custom plugin?
Just double checking, I think temporal batching is currently happening in DS 5.0.2
Ex: if I have a single source, set streammux batch-size to 2, and run the pipeline with GST_DEBUG=nvstreammux:7 , I see the following in the logs (TRACE statements removed)