2.You can remove the nvinfer in your plugin to check the fps
=> Yes, FPS rate is the same after removing nvinfer.
3.You can also use the trtexec --onnx=your_model.onnx to check the TensorRT perf first
=> I tried this. Result:
GPU Compute Time: min = 6.51514 ms, max = 8.5155 ms, mean = 6.65694 ms, median = 6.58362 ms
Log attached. trtexec_onnx_log.txt (22.7 KB)
Since FPS rate is the same after removing nvinfer. The model analysis is not a bottleneck. We currently do not have 120fps camera on hand. So could you assist in further analysing?
You can just use source and fakesink to test the fps first and increase the plugins gradually. Finally, we can investigate whether it is a bandwidth bottleneck or a software bottleneck. Thanks
How can I test the FPS using only source and fakesink? Should I disable each plugin in the deepstream-app config file, for example by setting enable=0 for [primary-gie], to run deepstream-app?
You can just use the cli below: gst-launch-1.0 v4l2src device=/dev/video0 ! fpsdisplaysink text-overlay=0 video-sink=fakesink sync=0 -v
Hi @yuweiw, I execute the “gst-launch-1.0 v4l2src device=/dev/video0 ! image/jpeg,format=MJPG,width=1920,height=1080,framerate=120/1 ! fpsdisplaysink text-overlay=0 video-sink=fakesink sync=0 -v” command and the result is as shown in "
fps.log (30.7 KB)". It seems that the fps can reach 120.
Could you advise me on how to proceed with the next test?
You can try to add nvstreammux, nvinfer and other plugins one by one to the pipeline. Observe the changes in frame rate during this process.
Hi @yuweiw, thanks for your command to check Camera capbility. As our problem mentioned at beginning, we need to open 2 CAMs(1080P@120fps) simulatneously. We used your fakesink checking command to open 2nd CAM and got error message “v4l2src0: Failed to allocate required memory.”
Do you have any idea to solve the issue?
Step1: Cam1: 1080P@120fps launching - OK!
Step2: Cam2: 1080P@120fps launching - can’t open successfully!
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
As @DaneLLL replied before about this problem, the bandwidth is the bottleneck of this memory problem.
Otherwise, the problem can be solved by reducing the frame rate or resolution.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.