Pixel distortion in miltiple rtsp input

Accelerated Computing Intelligent Video Analytics DeepStream SDK
26

I did what you said , now i can only get 200 fps using 10 same rtsp input !
If i add more i will get pixel distoration!
This is what i get after using 11 :(still only 50% gpu)

7d80693e42185a1fd8cb7e9996c37aed2f7a6971
7d80693e42185a1fd8cb7e9996c37aed2f7a69711831×863 384 KB

Nvurisrcbin->streammux->pgie->tracker->demux - -every camera- → queue → nvvideoconvert → nvosd-> nvvideoconvert-> caps-> encoder ->rtppay-> udpsink

Nvurisrcbin :
uri_decode_bin.set_property(“cudadec-memtype”, 0)
uri_decode_bin.set_property(“select-rtp-protocol”, 4)
uri_decode_bin.set_property(“rtsp-reconnect-interval”, 500)
#rtspsrc
if source_element.find_property(‘drop-on-latency’) != None:
source_element.set_property(“drop-on-latency”, True)
source_element.set_property(“latency”, 2000)
source_element.set_property(“is-live”, True)
source_element.set_property(“protocols”, “tcp”)
streammux
streammux.set_property(‘live-source’, 1)
streammux.set_property(‘gpu-id’, 0)
streammux.set_property(‘width’, 1920)
streammux.set_property(‘height’, 1080)
streammux.set_property(‘batch-size’, number_sources)
streammux.set_property(‘batched-push-timeout’, 40000)
nvinfer :
it’s correct
tracker:

tracker.set_property('tracker-width', 640)
tracker.set_property('tracker-height', 384)
tracker.set_property('gpu-id', 0)
tracker.set_property('ll-lib-file', "/opt/nvidia/deepstream/deepstream/lib/libnvds_nvmultiobjecttracker.so")
tracker.set_property('ll-config-file', "config_tracker_NvDCF_perf.yml")

mem_type = int(pyds.NVBUF_MEM_CUDA_UNIFIED)
nvvidconv.set_property(“nvbuf-memory-type”, mem_type)
nvvidconv.set_property(“gpu-id”,0)
nvvidconv_postosd.set_property(“nvbuf-memory-type”, mem_type)
nvvidconv_postosd.set_property(“gpu-id”,0)

nvosd.set_property(‘process-mode’,OSD_PROCESS_MODE)
nvosd.set_property(‘display-text’,OSD_DISPLAY_TEXT)
nvosd.set_property(“gpu-id”,0)

caps.set_property(
“caps”, Gst.Caps.from_string(“video/x-raw(memory:NVMM), format=I420”)
)
nvv4l2h264enc:
encoder.set_property(‘bitrate’, bitrate)
encoder.set_property(“gpu-id”,0)
rtph264pay:nothing

sink.set_property(“host”, “224.224.255.255”)
sink.set_property(“port”, updsink_port_num)
sink.set_property(“async”, False)
sink.set_property(“sync”, 1)sink.set_property(“host”, “224.224.255.255”)

27

You are not displaying locally but remotely. Please do not set “sync” to 1 with your udpsink. The UDP protocol do not guarantee packet lost.

28

Tt for responding
I still have that problem
The problem is about gpu process
It only get 50% even with packetloss i must get more fps
I can only get 200 fps in rtsp input and 1200 fps in video
It’s not correct !

29

As your description, you use 10 same rtsp input and get 200fps. If you add more, you get pixel distortion. There are two issues here, to use the same source repeatedly, some rtsp server(maybe your source is the ip camera) may have issue when the connection from the same client is too much. The second problem is that you can add more sources, the more sources will use more GPU. The only side effect is the pixel distortion which I’ve told you from the beginning that the distortion is caused by ethernet packet lost, it has nothing to do with GPU.

30

You may try 20 different ip cameras and set your sink element as nveglglessink to check whether there are still pixel distortion.

31

The problem are thease
By adding more rtsp gpu processing is still 50%!
Facedetect only gives us 200 fps when it can give us 1200 fps using videos!!
I’m not talking about pixel distoration becauze it happend when fps is too low cuze of my last two problem

32

I found the issue !!
I don’t know why but i when i use nvds_enable_component_latency … i found this:
After every demux src there is a muxxer with high latency !!
Why is it exist? And how can i fix it?
I even remove queue but nothings changed

33

Any help?

34

What is your model? Have you measured your model’s performance? The GPU loading is 50% means that the GPU is not the bottleneck of the pipeline.

What do you mean by this?

Have you monitored the GPU and CPU loadings when you run the case?

35

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