Weird artefacts occasionally appear in smart record videos

I use smart record to record small video clips when a detection occurs.

Occasionally I get a strange artefact where it looks like other streams are bleeding into the frame.

See this link for a good example.

This only occurs when I have smart record tapped into a tee AFTER splitting the streams again with the stream demuxer element and re-encoding (because I want to see the bboxes in the recordings).

My guess is it’s a bug in the stream demuxer.
When I tap in smart record before decoding - so that there is no stream demuxer used then I never see these strange artefacts.

At the time of testing I had 4 rtsp streams from hikvision cameras at 1080p and 25fps.
Pgie interval is 0.
Memory usage was around 50% and cpu utilisation was low (under 12%). But then around 7 hrs before this clip was recorded memory usage jumped to over 65%. This is very unusual because with 4 cameras it runs for months non-stop at around 50%.
A couple hours later and memory was over 95% used !!
It was around this time that the video clip linked here was recorded.
So something has caused a memory leak possibly. Once the jetson nano has then than about 7% memory available all sorts of weird things start to happen.

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) jetson nano
• DeepStream Version latest
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs) bugs
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing) you will need to create your own pipeline so that you use a stream demuxer. You don’t need smart record and can just view the footage with a tiler however it can be several hours before the issues occurs.
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

@jasonpgf2a Is there a easier way to reproduce this problem with deepstream-app?

Not possible to duplicate with deepstream-app because smart record needs to be tapped (tee’d) into the pipeline after processing (decode streammux nvinfer tracker streamdemux).
I’m happy to share my app though - Have added you to th git repo before so you could investigate a different issue.