Is there any way to custom adapter type of the advanced setting on DL Workbench.

[ty_s]

Hello.

Is there any way to add new adapter type to advanced setting on Open VINO DL Workbench ?

I would like to measure accuracy of YOLOX and the adapter type is not supported by the DL Workbench Version 2021.4.
I found YOLOX is supported by the later version but I have to use the Version 2021.4.

Is there any way to get accuracy of YOLOX using the DL Workbench Version 2021.4 ?

Thank you.

[Zulkifli_Intel]

Hello Ty_s

Thank you for reaching out to us.

To specify the accuracy configuration, you can refer to Accuracy Configuration. It demonstrates how to configure Basic Accuracy Settings and Advanced Accuracy Settings.

Is there any specific reason why you want to use OpenVINO 2021.4 instead of the latest release? Also, please provide us with the following detail:

• Yolox-tiny model (IR model or link to original model)
• Which OS did you use?

Sincerely,

Zulkifli 

[ty_s]

Hello Zulkifli
Thank you for the reply!

The reason for using OpenVINO 2021.4 is as requested by my customers.
They don't want to update OpenVINO from 2021.4 to the latest version, so I need to use version 2021.4.

The Link to original model is below and I use YOLOX-Tiny.
https://github.com/Megvii-BaseDetection/YOLOX/tree/0.2.0#light-models 

OS is Lubuntu 20.04.4

[Zulkifli_Intel]

Hello Ty_s,

You can change the adapter type by follow these steps:

• Import model
• Select environment
• Import annotated dataset
• Create project
• Once completed, click "Perform"
• Click "accuracy report
• Provide accuracy configuration
• Click Advanced to configure adapter type

Unfortunately, as of now, yolox-tiny model is not yet supported in DL Workbench. It will show an error when trying to create accuracy report.

For your information, yolox-tiny is officially supported by OpenVINO 2022.1. Where the model can be downloaded from Intel Open Model Zoo. You can use the Deep Learning accuracy validation framework to get the accuracy of the yolox-tiny model.

Additional information, starting with the Intel Distribution of OpenVINO toolkit 2021.3 release, DL Workbench is available only as a prebuilt Docker image. Reference to DL Workbench is kept in OpenVINO installation but now pulls the pre-built image from Docker Hub instead of building it from the package. Please refer to Release Notes for OpenVINO 2021 for further detail.

Sincerely,

Zulkifli 

[Zulkifli_Intel]

Hi Ty_s,

This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question.

Sincerely,

Zulkifli