Hyperspectral image with DL workbench

[ShaLu]

Hi,

I am working with hyperspectral images that usually have more than 3 channels. To import my model into the DL workbench, there are two domains that can be chosen, CV or NLP.  If I select the CV model, it is required to choose RGB or BGR. As my model has 6 channels, no matter either one I choose, MO fails.

Is there any possible way to make hyperspectral images work with the DL workbench?

Thanks in advance!

Regards,

Sha

[Peh_Intel]

Hi Sha,

In this case, please try to import OpenVINO IR model directly instead of converting model to IR through DL Workbench. Select ‘CV’ for Domain and ‘OpenVINO IR’ for Framework.

For your information, you can install OpenVINO Development Tools through the PyPI repository. To install and configure the components of the development package for working with specific frameworks, use the pip install openvino-dev[extras] command.

For example, to install and configure the components for working with all frameworks, use the following command:

pip install openvino-dev[caffe,kaldi,mxnet,onnx,pytorch,tensorflow2]

Next, convert the model into Intermediate Representation (IR) with Model Optimizer by using the console script, mo.

Regards,

Peh

[ShaLu]

Hi Peh,

Thanks for your reply. 

Following the step you instructed, I have successfully imported my model into the DL workbench. Here I have another question. How could I see the inference results? I have tried to click CreateAccuracy Report and Visualise Output, but neither succeeded. When I click CreateAccuracy Report, I got the error information: 'This report is not available for projects with not annotated dataset.Import annotated dataset to create Accuracy Report'. For Visualise Output, the hyperspectral image cannot be imported because it has 6 bands. 

Is there any way that I could see the inference results?

Regards,

Sha

[Peh_Intel]

Hi Sha,

After importing model and unannotated data, and create Project, the Projects page will open automatically with the analyzed performance (FPS).

To create Accuracy Report, you have to upload annotated data. Please refer to Dataset Types for the details of the formats.

Regards,

Peh

[Peh_Intel]

Hi Sha,

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

Regards,

Peh