Intel® Edge Software Hub
Get answers from community peers to your questions about building Edge Software Hub solutions for edge compute node.
Announcements
Welcome to the Intel Community! If you like the answer to your question, please mark it as 'Accepted Solution' to help others.

The Edge Software Vision Package for Red Hat Enterprise Linux is now available here.

Learn more about developing on Intel® Hardware and Software here.
389 Discussions

Is there a relationship between the camera's Frame Size (640*640) and CPU & Memory usage in EII?

Louis8
New Contributor I
670 Views

 

I ran the test as below:

I am continuously running EII 4.1.0 and checking the camera speed. In the meantime, I changed the camera's frame size to 640*640 for inference and tested it.
This clearly shows good performance when the existing frame size is not changed.

The environment I tested was with only one USB camera installed. I'm not sure what problems will arise when connecting to multiple cameras.

 

Can I reduce CPU and memory usage on an EII 4.1.0 system if I adjust the frame size to a smaller size?

docker_stats_graph.png

 

The environment I tested is as follows.

1. Use cases: video-streaming-evam.yml

2. System Hardware Specification:

  • Processor: 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz 2.80 GHz
  • Installed RAM: 8.00GB (7.68GB usable)
  • System Type: 64-bit operating system, x64-based processor
Labels (1)
0 Kudos
7 Replies
Louis8
New Contributor I
633 Views

In this test, we are using two camera pipelines of the Microservice.

The cameras are regular USB camera. We adjusted the camera's frame size and tested it at 640 x 640. The test lasted about 2 days.

 

The table below shows some of the extracted usage data.

스크린샷 2023-11-30 163331.png

 

Inline-image-2023-11-29 09.23.22.065.png

Most of the CPU is occupied by the Visualizer. And Memory usage was up to 1.427 GiB by Visualizer.
If it continuously refresh the Chrome browser, the CPU usage will increase and the PC will eventually lose its senses.

How could we overcome this problem?

0 Kudos
JesusE_Intel
Moderator
620 Views

Hi Louis8,


Thank you for reporting! Let me open a ticket with the development team for further assistance. I also noticed this behavior on my system when running the video-streaming-evam use case with the following pipeline. Could you please share your pipeline?


"pipeline": "v4l2src device=/dev/video0 name=source ! video/x-raw,format=YUY2,width=640,height=480 ! videoconvert ! video/x-raw,format=BGR ! appsink name=destination"


Regards,

Jesus


0 Kudos
Louis8
New Contributor I
614 Views
The values of the pipeline are as follows.
 
"pipeline": "v4l2src device=/dev/video0 name=source ! video/x-raw,format=YUY2 ! videoconvert ! video/x-raw,format=BGR ! appsink name=destination",
0 Kudos
Louis8
New Contributor I
605 Views

Hi, Jesus

 

I share another test data here.
This test is the result of a test that only observed the web browser without refreshing it.

You can see that the memory occupancy is very constant.

It seems that the browser was dead between 00:00 AM and 6:00 AM. So, Visualizer's CPU usage is almost non-existent during that time.
After refeshing once, I could see the CPU taking up up to 160% again. But the numbers didn't rise any further.


Below are the test conditions.

AppContexts:
- ConfigMgrAgent
- EdgeVideoAnalyticsMicroservice/eii
- Visualizer/multimodal-data-visualization-streaming/eii
- USB Camera 2ea (Multi Instance)
- Frame size 640 x 640

 

docker_stats_graph-no-refreshing.png

0 Kudos
Louis8
New Contributor I
514 Views

I have conducted the test once again. It is EII version 4.1.0. For the test conditions, I limited the time of EVAM's Publisher to 0.1 seconds and conducted the test. However, the issue still persists where the Web Visualizer's CPU usage does not drop once it goes up. Is there a good solution for this problem?

Below are the test conditions.

AppContexts:

  • ConfigMgrAgent
  • EdgeVideoAnalyticsMicroservice/eii
  • Visualizer/multimodal-data-visualization-streaming/eii
  • USB Camera 2ea (Multi Instance)

 

My test results are as follows. Unlike the previous test, the frame size of the camera remained the same as the original source.

CPU and Memory Usage Over Time for Visualizer and EVAM (v4.1.0) - Publisher(0.1s delay) - 2 Cams.png

0 Kudos
JesusE_Intel
Moderator
428 Views

H Louis8,


Looks like your system does not meet the minimum requirements for Edge Insights for Industrial. Could you try to test on a different system or increasing the memory?


Getting Started — EII Documentation v1.0 documentation (intel.com)

System Requirement

Details

Processor

8th generation Intel® CoreTM processor onwards with Intel® HD Graphics or Intel® Xeon® processor

RAM

Minimum 16 GB

Hard drive

Minimum 128 GB and Recommended 256 GB

Operating system

Ubuntu 20.04


Regards,

Jesus


0 Kudos
Louis8
New Contributor I
418 Views

Of course. I will test it like that.

Below are the test results with camera encoding and decoding removed.

Please refer to it for work.

CPU usage is limited, so there doesn't seem to be any major problems.

Thanks in advance.

 

BR,

Louis

docker_stats_graph-docker_stats-no_encoding-4.1.0.png

 

0 Kudos
Reply