Getting only 20fps on python deepstream app on 6.3 on a 30 fps camera

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) DGPU NVIDIA GeForce GTX 1050
• DeepStream Version 6.3.3
• TensorRT Version 8.6.1.6-1+cuda12.0
• NVIDIA GPU Driver Version (valid for GPU only) 530.41.03
• Issue Type( questions, new requirements, bugs) Performance
• 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)
Using deepstream-test1-usbcam app on my laptop with deepstream installed, no config changes
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)
I am trying the deepstream-test1-usbcam on my ubuntu laptop running deepstream and getting only 20FPS PERF for my webcam that can do 30FPS.
I tried using the performance tuning tips of live source, specify framerate on the capsfilter element, but nothing seems to take it more than 20FPS. What else can I check?

➜  deepstream-test1-usbcam git:(master) ✗ v4l2-ctl -d /dev/video0 --list-formats-ext                                                                                                   
ioctl: VIDIOC_ENUM_FMT
	Type: Video Capture

	[0]: 'MJPG' (Motion-JPEG, compressed)
		Size: Discrete 1280x720
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 960x540
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 848x480
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 640x480
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 640x360
			Interval: Discrete 0.033s (30.000 fps)
	[1]: 'YUYV' (YUYV 4:2:2)
		Size: Discrete 640x480
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 640x360
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 424x240
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 320x240
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 320x180
			Interval: Discrete 0.033s (30.000 fps)
		Size: Discrete 160x120
			Interval: Discrete 0.033s (30.000 fps)
2

Please refer to the performance FAQ on whether some items can be applied. Besides, please also check whether GPU usage (nvidia-smi dmon).

4

Tried the applicable recommendations on the performance FAQ page. Can I try anything more?

  1. I could not find instructions on how to measure latency in python apps. So could not enable measurement for NVDS_ENABLE_LATENCY_MEASUREMENT. The environment variables do nothing on a python app.
  2. This is on a laptop so cannot move GPU slots
  3. Set batched-push-timeout to 30 in streammux
  4. height and width of streammux set to 640 X 480
  5. qos set to 1 for sink.
    I could not find all other recommendations applicable to this set up.

Here is nvidia-smi output when app is running:

# gpu    pwr  gtemp  mtemp     sm    mem    enc    dec   mclk   pclk 
# Idx      W      C      C      %      %      %      %    MHz    MHz 
    0      -    52      -    23     11      0      0   3504   1290 
    0      -    52      -    24     11      0      0   3504   1290 
    0      -    52      -    21     10      0      0   3504   1290 
    0      -    53      -    22     10      0      0   3504   1290 
    0      -    52      -    22     11      0      0   3504   1290 
    0      -    53      -    20      9      0      0   3504   1290 
    0      -    53      -    21     10      0      0   3504   1290 
    0      -    53      -    20      9      0      0   3504   1290 
    0      -    53      -    19      9      0      0   3504   1290
5
  1. NVDS_ENABLE_LATENCY_MEASUREMENT is for C code, please refer to DeepStream SDK FAQ - #12 by bcao 4 to measure the latency of the pipeline components.
  2. if testing performance, please replace nv3dsink with fakesink.
  3. please set nvinfer’s interval to improve performance. please find interval in nvinfer.
6

Thank you for the tips.

  1. The DeepStream SDK FAQ - #12 by bcao 4 link you shared seems to be for C code. Is there similar guideline for python?
  2. Replaced with nveglglessink - same performance
  3. nvinfer interval changed to max of 24, still same performance :(

Is there anything else I can check?

7

Finally figured it out - it turned out to be a camera issue.
Used this video to fix it https://www.youtube.com/watch?v=iuic3po8kvQ&ab_channel=codergopher

1 Like
8

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