Please provide complete information as applicable to your setup.
Good day
I am using deepstream along with the python bindings to process streams. All is going well, but the GPU of my computer is consistently crashing when I try to test a certain feature of the program. I suspect it is due to the processing constraints of the GPU-it crashes when the processing requirements exceeds it capabilities.
Is there a way for me to decrease the rate at which images are processed in the pipeline(Like configuring the streammux plugin to release frames at a slower rate)? In believe that this would prevent the GPU from crashing.
So effectively this section of the pipeline would look like this:
… → streammux → queue → videorate ->queue → …
The videorate plugin does not seem to have an influence. The video is able to be processed, but the frame rate in the processed video is still 60 FPS. I am streaming a video that is stored on my local device.
When I try to set the framerate parameter then I get this message:
Traceback (most recent call last):
File “main.py”, line 58, in
sys.exit(main())
File “main.py”, line 28, in main
pipeline, streammux, tiler, nvosd, ALL_BINS = construct_pipeline(sources, ip_addreses, ONNX, fileNames)
File “/home/aizatron/repos/RnD-video-tracking/src/pipeline_setup.py”, line 161, in construct_pipeline
videoRate.set_property(“framerate”, 10/1)
TypeError: object of type GstVideoRate' does not have property framerate’
Yes, that is because framerate is not supposed to be written using set_property().
if you run “gst-inspect-1.0 videorate”, you will not find framerate in the element properties
your code respective pipeline on a terminal would look lik this ... ! videorate framerate=10/1 ! ...
but it actually works like this ... ! videorate ! "video/x-raw,framerate=10/1 ! ...
so you can change your code to something like this…
Method 1 : (NOTE: if you have nvstreammux right after the videorate element, you can do the linking of srcpad and sinkpad of this capsfilter and nvstreammux)