Deepstream 6.4 and very low FPS

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Working with DS.6.4, dGPU. Driver Version: 535.183.06 CUDA Version: 12.2.
I have a pipeline with multiple rtspsrc input bins that goes into nvstreammux, than further to nvinfer ! nvdsosd ! nvstreamdemux ! fakesink.

The camera input configured to 30fps, 1920x1080.

The issue is that the FPS drops depending on the count of input streams.
If running 2 inputs, the FPS is normal. If running 1 or 3 streams, the FPS drops significantly.

Please, help with issue as I have no idea of where the issue might be.
Thanks in advance!

3

Please tell us what is your dGPU?

Please refer to DeepStream SDK FAQ - Intelligent Video Analytics / DeepStream SDK - NVIDIA Developer Forums to make sure the nvstreammux parameters are correct for the multiplr RTSP streams.

4

GPU: NVIDIA GeForce RTX 2080 Ti.
streammux parameters: live-source=1, batched-push-timeout=4000000 batch-size=1 (depending on number of input streams).

I started debuging the pipeline to find out which is the main element that drops the FPS. Still in search.

But now i reduced the pipeline to gst-launch-1.0 -e rtspsrc location="rtsp://ip/axis-media/media.amp" protocols=tcp ! rtph264depay ! h264parse ! queue ! decodebin name=decoder ! nvvideoconvert ! 'video/x-raw(memory:NVMM),format=(string)NV12,width=(int)1920,height=(int)1080' ! smux.sink_0 nvstreammux name=smux width=1920 height=1080 batch-size=1 ! fakesink.

Still with this reduced pipeline, the problem exists. It looks like the only one possible is the nvstreammux.

5

MUXER_BATCH_TIMEOUT_USEC = 20000 is the only value I can use to keep stream state normal.

6

Please set the live stream related parameters with nvstreammux.

gst-launch-1.0 -e rtspsrc location="rtsp://ip/axis-media/media.amp" protocols=tcp ! rtph264depay ! h264parse ! queue ! decodebin name=decoder ! nvvideoconvert ! 'video/x-raw(memory:NVMM),format=(string)NV12,width=(int)1920,height=(int)1080' ! smux.sink_0 nvstreammux name=smux width=1920 height=1080 batch-size=1 live-source=1 batched-push-timeout=33000 ! fakesink
7

Thanks, nvstreammux live-source already enabled. That did not fixed the problem.

8

Have you set the correct “batched-push-timeout” value?

9

Previously i had batched-push-timeout=20000. It helps a lot.
batched-push-timeout=33000 also seems to be good.

10

Please refer to DeepStream SDK FAQ - Intelligent Video Analytics / DeepStream SDK - NVIDIA Developer Forums

11

With the new nvstreammux, the stream hangs.

Hangs if fakesink async=1 or async=0.

After timeout it reports : 3671890 0x7fffcc00f100 WARN rtspsrc gstrtspsrc.c:3638:on_timeout_common: source 104ddf0e, stream 104ddf0e in session 0 timed out

12

Please check the RTSP server side.

13

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks