About EII ALPR replacement custom model and replacement camera src and replacement intel GPU issues

[jacky0327]

Hi Intel,
I'm currently experimenting with EII ALPR on Ubuntu 20.04,
https://www.intel.com/content/www/us/en/developer/articles/reference-implementation/automated-license-plate-recognition.html

I have successfully set up this case on my computer with the operating system Ubuntu 20.04.


I would like to ask if I can change the two models of ALPR by myself?
I found two models at the following path:

/home/gvsai/automated_license_plate_recognition/Automated_License_Plate_Recognition_2022.1/Automated_License_Plate_Recognition/EII-LicensePlateRecognition-UseCase/config/VideoIngestion/udfs/python

• license-plate-recognition-barrier-0007
• vehicle-license-plate-detection-barrier-0106

Can I change to the other models of openvino model zoo?
https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/intel

In addition, I would like to ask how to change the camera src from local files to webcam via rtsp/tcp?
I refer to the following urls:

and

https://github.com/open-edge-insights/video-ingestion/blob/master/docs/gva_doc.md

Finally, I would like to ask how to change the inference to use Intel's built-in GPU?
https://github.com/open-edge-insights/eii-core#task-2--install-prerequisites

Thank you Intel for providing EII, please help me solve the above problems, thank you very much.

[JesusE_Intel]

Hi jacky0327,

The Automated License Plate Recognition Reference Implementation is a great example on how OpenVINO models can be used to extract the license plates from a video file. It's definitely possible to take this reference implementation further and use a different model, device and video source.

Let me run some quick tests and give you more specific information on how to accomplish this. In the meantime, I recommend you look at the application configuration files located under <INSTALL_PATH>/automated_license_plate_recognition/Automated_License_Plate_Recognition_<version>/Automated_License_Plate_Recognition/EII-LicensePlateRecognition-UseCase

To inference to a GPU device, you would need to specify "GPU" as the target device.

For RTSP cameras, I recommend starting with the Video Ingestion documentation here:

https://open-edge-insights.github.io/IEdgeInsights/VideoIngestion/docs/rtsp_doc.html?highlight=rtsp#rtsp-camera

Regards,

Jesus

[jacky0327]

Hi JesusE_Intel,

Thank you for your reply. We really need to try to replace other models in EII ALPR to evaluate the usability of EII

We are also developing ALPR products.

Regarding the quick tests you mentioned, please teach me if you have any progress. Thank you very much, our goal is change to our custom model.

Regarding chang the GPU to the specified device, where can I make the change, and what are the steps?
Is it to enter the docker to make changes after the docker is set up? Or make the changes before the docker images built?

Thank you for providing the reference of RTSP Cam, I have the same questions as above.

Regards,
Jacky

[JesusE_Intel]

Hi Jacky,

Quick update: I started configuring the reference implementation to use the SafetyGear UDF example and model from EII. I was able to update the WebUI to display safety gear UDF. However, the ia_video_ingestion_alpr container is failing to start. I've reached out to my peers for further assistance.

https://github.com/open-edge-insights/video-custom-udfs/tree/master/PySafetyGearAnalytics/safety_gear

Regards,

Jesus

[jacky0327]

Hi JesusE_Intel,

Thanks for your quick update, I also tried to update the config of UDF, the same as you, ia_video_ingestion_alpr cannot get start.

Did you make changes to the config before the webui make started?

Regards,
Jacky

[jacky0327]

Regards,
Jacky

[JesusE_Intel]

Hi Jacky,

I'm really sorry about the delay from our side. I escalated this to the development team. Let me follow up with them to get the latest update.

I appreciate your patience.

Regards,

Jesus

[JesusE_Intel · Accepted solution]

Hi Jacky,

Sorry for the delay in our response, the development team has provided the following feedback.

We don't have a guide to make changes or customize this reference implementation. However, it's possible to make changes by modifying the config files and rebuilding the container. Here are some general steps:

cd /opt/intel/eif/IEdgeInsights/VideoIngestion_ALPR

edit config.json. here edit the model and add RTSP (additional info below)

cd /opt/intel/eif/IEdgeInsights/build

chmod +x builder.py

./builder.py

docker-compose -f docker-compose-build.yml build (This will rebuild all the container services)

For this specific reference implementation, it may not be possible to use an RTSP stream since the Web UI only includes browse for local video files. You can try to modify the config.json file and add source parameters and change the pipeline to your RTSP stream path. For example: Advanced Guide — EII Documentation v1.0 documentation (intel.com)

Hope this helps!

Regards,

Jesus

[JesusE_Intel]

If you need any additional information, please submit a new question as this thread will no longer be monitored.

[jacky0327]

Hi JesusE_Intel,

Thank you very much for your reply, ︀If we have needs in the future, ︀we will come back to refer to this Case again.

Regards
Jacky