Yolov5 run slow with python binding

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform :Jetson Nano
*• DeepStream: 6.0
• JetPack Version: 4.6.1
• TensorRT Version: 8.2.1.8
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type: questions

Hi everyone I use this guide :

https://gist.github.com/bharath5673/800a18cc7474ce9c22fda6deaaa98354#comments

To run yolo5s on my jetson nano, When I run by the command:

deepstream-app -c deepstream_app_config.txt

I get result around 8.5 FPS

and when I try using binding in python I get very slow result I can’t understand why? of course I need to mention that I new with deepstream I learn this only this week and using python binding can very help me to my goals.

I add to short videos of the result I would be happy to any help thank you :)


3

from your screenshot, using python sample, the fps also is 8, what do you mean about “very slow result”?

4

Hi first thank you to reply, I didn’t understand how you see in the screenshot 8 fps maybe I miss something but in general, what I mean is that you can see that the video run more slowly on the screen when I use python binding, In addition, you can see in the video that I get in the terminal the message There may be timestamping problem or this computer is to slow. and pay attention to NVDEC there are different between videos.

5

from the second video, I can see the video is stuck, where did you get that deeptream-yolo.py? it is custom code? you can set sync=0 in [sink] configuration.

6

I get deepstream-yolo.py from the same git above, I use in the same configuration.txt so I really don’t understand what happen.

note (10.):

https://gist.github.com/bharath5673/800a18cc7474ce9c22fda6deaaa98354#comments

and when I run yolo-Tracker.py I get the same slow result absolutely I do something wrong maybe configuration or something like that I just need to figure out.

7

hi is this the configuration? Deepstream/deepstream_app_config_yolo.txt at main · bharath5673/Deepstream · GitHub
please set num-sources to 1, which is similar to C version 's configuration.

8

Hi sorry I despaired, I try to run my python code with your configuration (change num-source to 1) and get this error

Creating Pipeline 
 
Creating streamux 
 
Creating source_bin  0  
 
Creating source bin
source-bin-00
Creating Pgie 
 
Creating tiler 
 
Creating nvvidconv 
 
Creating nvosd 
 
Creating transform 
 
Creating EGLSink 

Could not find group property
** ERROR: <gst_nvinfer_parse_config_file:1303>: failed
Adding elements to Pipeline 

Linking elements in the Pipeline 

Now playing...
1 :  file:///opt/nvidia/deepstream/deepstream/samples/streams/sample_1080p_h264.mp4
Starting pipeline 


Using winsys: x11 
0:00:00.298257623 11860   0x558effe290 WARN                 nvinfer gstnvinfer.cpp:794:gst_nvinfer_start:<primary-inference> error: Configuration file parsing failed
0:00:00.298301739 11860   0x558effe290 WARN                 nvinfer gstnvinfer.cpp:794:gst_nvinfer_start:<primary-inference> error: Config file path: deepstream_app_config_yoloV5_check.txt
Error: gst-library-error-quark: Configuration file parsing failed (5): /dvs/git/dirty/git-master_linux/deepstream/sdk/src/gst-plugins/gst-nvinfer/gstnvinfer.cpp(794): gst_nvinfer_start (): /GstPipeline:pipeline0/GstNvInfer:primary-inference:
Config file path: deepstream_app_config_yoloV5_check.txt
Exiting app

In general I run my code with this configuration file:
config_infer_primary_yoloV5.txt (629 Bytes)

I am really sure that I do something wrong with the configuration and because of this I get a slow result I will be very happy if you can help me to figure it out.

9

** ERROR: <gst_nvinfer_parse_config_file:1303>: failed
here is an error, parsing configuration file failed, can you narrow down this issue? you can compare it with original file.

10

Ok I try to compare and not see really different still get error parsing when I use this configuration:
deepstream_app_config_yoloV5_check.txt (4.6 KB)

When I run yolo with the configuration:
config_infer_primary_yoloV5.txt (629 Bytes)
I don’t get parsing problem but my yolo run very slow like you see in the short video.
note that the parser error is: Could not find group property

There is some log or something that I can add to the original code to understand where is the parsing problem?
I control my jetson nano with remote ssh using pycharm and try to debug the problem but still can’t figure out what happened I came to the line code
pgie.set_property('config-file-path', "deepstream_app_config_yoloV5_check.txt")
and just get the error parsing it’s difficult to understand where it falls and I try to compare it to another deepstream-config.txt it seems to me ok.

11

The solution is from your forum I have a problem of dropped frames.

https://forums.developer.nvidia.com/t/error-a-lot-of-buffers-are-being-dropped-when-running-ds-sdk-python-sample-ipcamera-on-nano/110886/5

Hi,
Please apply the modification to deepstream_test_3.py and try again.
1 set ‘live-source=1’ to streammux

streammux.set_property('live-source', 1)

2 Replace nveglglesink with nvoverlaysink and set ‘sync=0’

- transform=Gst.ElementFactory.make("nvegltransform", "nvegl-transform")
+ transform=Gst.ElementFactory.make("queue", "queue")

- sink = Gst.ElementFactory.make("nveglglessink", "nvvideo-renderer")
+ sink = Gst.ElementFactory.make("nvoverlaysink", "nvvideo-renderer")
+ sink.set_property('sync', 0)

3 Apply the setting of Nano to config file
https://devtalk.nvidia.com/default/topic/1058597/deepstream-sdk/-nano-deepstream-test3-app-not-working-as-expected-for-multiple-video-source/post/5368352/#5368352

12

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