Fail to run the Social Distancing application on GPU device

[chiao]

Hi Intel,

I am testing the below " Social Distancing for Retail Settings Reference Implementation" , but facing the issue with running the application on GPU device.

Could you help us to solve the issue?? Thanks!!

Please see the attached file for the logs.

https://www.intel.com/content/www/us/en/developer/articles/reference-implementation/social-distancing-for-retail-settings.html

BR,

Denise

[ChunSheong_F_Intel]

Hi Chiao,

Does the RI work when you are running on CPU? Which platform/hardware/CPU are you currently using?

[chiao]

Hi ChunSheong,

It worked well on CPU.

We use intel NUC i5 for testing, thank you!!

BR,

Denise

[Xiang_Intel · Accepted solution]

Hi Denise,

May I know if I can get to know the CPU platform from you so that I can better help in resolving the issue?

In the meantime can you help to change this line in the Dockerfile in the following directory and compile the docker container and run it again?

1. Edit OpenVINO dockerfile to latest version tag

In Social_Distancing_for_Retail_Settings_<version>/Social_Distancing_for_Retail_Settings/retail-settings, open the Dockerfile, and edit line 13 from

FROM openvino/ubuntu20_data_runtime:2021.3 -> FROM openvino/ubuntu20_data_runtime:2021.4.2

2. Add the following line into requirements.txt

itsdangerous==2.0.1

3. Compile the container again

sudo -E docker-compose build

4. Launch with docker-compose

export HOST_IP=$(hostname -I | cut -d' ' -f1)

sudo -E docker-compose up -d

Do note that, when you first launch the app with GPU, it does take a while for the GPU device to initialize, this is due to the first time bringing up of GPU in the container. So you will need to wait for a few minutes before you can see the result in Grafane as compared to launch with CPU.

Do let me know if you still have the issue.

Thanks,

XY

[chiao]

Hi XY,

Following are the CPU info for your reference :

CPU Info : 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
CPU Core Count : 8

After applying the adjustment you provided, it can work on GPU successfully.

But the stream display would show a black screen and the screen stops after about ten minutes.

Please help to check the log files, thanks!!

BR,

Denise

[Xiang_Intel]

Hi Denise,

This RI actually does not support inferencing on GPU devices which is stated in the documentation.

The fix I have provided was to help you to experience and run it on the GPU devices. Currently, I have tested the RI on my personal machine but I am unable to replicate your issue on my end.

For your info, since the video stream is started on a video file, each time the video is looping. There will have a black screen (which is the initialization process for the video looping). At the same time, I would suggest you have more RAM allocated to the machine if your machine has less than 8GB RAM.

Thanks and regards,

Lim Xiang Yang