Fail to run deepstream5.0 python sample app using anaconda python environment

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform: GPU
• DeepStream Version: 5.0
• JetPack Version: None
• TensorRT Version: 7.0.0.11
• NVIDIA GPU Driver Version: 440.33.01

• CUDA Version: 10.2
• CUDNN Version: 7.6.5
• Ubuntu: 18.04

Hi, please find my setup above. I am able to run the deepstream python sample app on my dGPU setup and see the popup window showing the inference results video. The command I use is shown as below:
python3 deepstream_test_1.py /opt/nvidia/deepstream/deepstream/samples/streams/sample_720p.h264
However, I cannot run this sample app in my anaconda python environment. Since I am in conda environment, I use the following command:
python deepstream_test_1.py /opt/nvidia/deepstream/deepstream/samples/streams/sample_720p.h264
Then I got:

ds_error
ds_error1267×806 96 KB

I have tried to copy the pyds.so file under the /opt/nvidia/deepstream/deepstream/sources/python/bindings/x86_64 to /my-path-to-anaconda3/lib/python3.6/site-packages/ and modify the deepstream_test_1.py to make sure import pyds from the anaconda environment.
Then I got:

ds_error2
ds_error21198×407 177 KB

I have also tried to copy the libnvdsgst_inferserver.so, libgstcoreelements.so, libnvdsgst_osd.so to /my-path-to-anaconda3/lib/python3.6/site-packages/, but it only helps to get rid of the first GStreamer WARNING. The rest two warnings remains.

Could you help me resolve this? Thanks in advance.

3

You do not need to copy the libnvdsgst_inferserver.so, libgstcoreelements.so, libnvdsgst_osd.so to /my-path-to-anaconda3/lib/python3.6/site-packages/, which GPU you are using? deepstream required nvidia GPU display card for output using type 2 eglglessink, if you use Tesla series card, you could follow this wiki to setup virtual display, Deepstream/FAQ - eLinux.org, or you could change python code sink from nveglglessink to fakesink.
#sink = Gst.ElementFactory.make(“nveglglessink”, “nvvideo-renderer”)
sink = Gst.ElementFactory.make(“fakesink”, “nvvideo-renderer”)
another option is you could use deepstream-test1-rtsp-out/ for rtsp streaming, then you could view the output through player vlc.

4

I am using RTX2080Ti.
And my problem is that I can run the test in native python3 environment but I cannot run the test in anaconda3 python3 environment. I am not sure if display is the cause of that.

5

Oh, missed you running from conda environments, so the issue related with conda environments.

6

Yes, I guess there is something wrong with my conda environment. The conda cannot find the gst or something related. Could you help me resolve this?

7

I do not have experience of conda, you may need to figure out yourself.
by google search or some other means.

9

Hi, I meet same problem as you, have you solved the problem already?

10

Hi,
no I gave up. I just use native environment

11

Would you mind sharing how you did all this setup in the local machine ? Like the commands that you have executed? I am having hard time setting it up