Deepstream Pipeline with branching

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson
• DeepStream Version 7.1
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, 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)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

I am working on deepstream pipeline where I have different types of games and each games can have multiple streams. I have to run a pipeline that have different models pgie, sgie for each game with multiple streams and then all the games branches merge into a single branch and run another sgie2 model and then tiler, nvosd, and sink
Branch A: Sources → Streammux → PGIE → SGIE –

Branch B: Sources → Streammux → PGIE → SGIE –

Both branches will merge into
Streammux → SGIE2 → Tiler → Sink

I just want to clarify if this is correct flow and another question is If there are three streams in branch A and two in branch B SGIE2 will run on all 5 streams?

3

No, it’s not the correct flow. You can refer to our sample code deepstream_parallel_inference_app to see if that meets your needs.

4

Can I add another model after merging? may be after meta-mux element

5

Yes. You can modify our source code to try that.

6 
gst-launch-1.0 \
nvurisrcbin uri=file:///opt/nvidia/deepstream/deepstream-7.0/samples/streams/sample_720p.mp4 name=src1 ! queue ! mux1.sink_0 \
nvurisrcbin uri=file:///opt/nvidia/deepstream/deepstream-7.0/samples/streams/sample_walk.mov name=src2 ! queue ! mux1.sink_1 \
nvurisrcbin uri=file:///opt/nvidia/deepstream/deepstream-7.0/samples/streams/sample_walk.mov name=src3 ! queue ! mux1.sink_2 \
nvurisrcbin uri=file:///opt/nvidia/deepstream/deepstream-7.0/samples/streams/sample_run.mov name=src4 ! queue ! mux1.sink_3 \
nvurisrcbin uri=file:///opt/nvidia/deepstream/deepstream-7.0/samples/streams/sample_walk.mov name=src5 ! queue ! mux1.sink_4 \
nvurisrcbin uri=file:///opt/nvidia/deepstream/deepstream-7.0/samples/streams/sample_walk.mov name=src6 ! queue ! mux1.sink_5 \
nvstreammux name=mux1 batch-size=6 width=1280 height=720 batched-push-timeout=40000 live-source=0 ! \
tee name=t \
	t. ! queue ! nvstreamdemux name=demux \
		demux.src_0 ! queue ! mux2.sink_0 \
		demux.src_1 ! queue ! mux2.sink_1 \
		nvstreammux name=mux2 batch-size=2 width=1280 height=720 batched-push-timeout=40000 live-source=0 ! \
		nvinfer config-file-path=pgie_yolov8.txt !
		nvvideoconvert ! 'video/x-raw(memory:NVMM), format=RGBA' ! metamux.sink_0 \
	t. ! queue ! nvstreamdemux name=demux1 \
		demux1.src_2 ! queue ! mux3.sink_0 \
		demux1.src_3 ! queue ! mux3.sink_1 \
		nvstreammux name=mux3 batch-size=2 width=1280 height=720 batched-push-timeout=40000 live-source=0 ! \
		nvinfer config-file-path=pgie_yolov8.txt !
		nvvideoconvert ! 'video/x-raw(memory:NVMM), format=RGBA' ! metamux.sink_1 \ 
	t. ! queue ! nvstreamdemux name=demux2 \
		demux2.src_4 ! queue ! mux4.sink_0 \
		demux2.src_5 ! queue ! mux4.sink_1 \
		nvstreammux name=mux4 batch-size=2 width=1280 height=720 batched-push-timeout=40000 live-source=0 ! \
		nvinfer config-file-path=pgie_yolov8.txt !
		nvvideoconvert ! 'video/x-raw(memory:NVMM), format=RGBA' ! metamux.sink_2 \
nvdsmetamux name=metamux active-pad=sink_0 ! \
nvinfer config-file-path=pgie_yolov8.txt ! \
nvmultistreamtiler rows=1 columns=2 width=1280 height=720 ! nvvideoconvert ! nvdsosd ! nv3dsink sync=false

I have created this pipeline based on deepstream parallel inference app. I have added nvvideoconvert and caps to convert frame to RGBA after nvinfer in every branch, so I can get the current frame from all the soruces since metamux only output one pad (or branch) with the metadata of all other branches.

My question is after merging the branches using metamux, will it have the actual source-ids for frames?

7

What do you mean the actual source-ids for frames? Did you encounter any problems when using this sample?

8

I haven’t tested the c++ application. I have just developed this gst cmd pipeline.

my question is if I add the pad_probe after metamux in python pipeline. It will have a frame_meta object, and frame_meta will have source_id. source_id will be in [0,1,2,3,4,5] since I have 5 different sources. or it will be a single source_id due to merge in metamux.

9

It will be in [0,1,2,3,4,5].

10

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

11

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