CNN classification using VAD.

[Deepak_C_R]

Hi,

I have created .graphml file using ModelOptimizer. I have loaded that to VAD tool also.

Now I need to load the trained model files into VAD, and would like to do classification of images ?.

Is it possible to do that in VAD ?, If yes can you provide the method to do the same ?
 

[Stav_S_Intel]

Hi Deepak,

No, VAD is used for helping graph level OpenVX applications and less so for Machine Learning. 

You cannot train model files in VAD. you can use Inference Engine and MO... 

Regards,

Stav

[Deepak_C_R]

Hi Stav,

I don't want train any thing using VAD.

I just wanted to load model files and graph ml file to VAD and predict an image.

[Stav_S_Intel]

Deepak,

This is not supported.

You can look at the OpenVX Graph created by MO to better understand the structure of the OpenVX Graph that was optimized. you cannot load model files as VAD is not meant for DL. it is not recommended to use VAD to re-generate the code you got from MO, as it will be generated without DL functionality. 

Regards,

Stav

[Deepak_C_R]

Thank you Stav.

[Deepak_C_R]

One more doubt.

In VAD I have seen activation, convolution, pooling, normalization layers. For what we are using those layers ?.

If we are not using VAD for DL, do we require that in VAD ?.

[Stav_S_Intel]

Hi,

We represent the nodes so you can review the optimization and understand the OpenVX Graph created, but you are not supposed to use them for DL Capabilities. VAD is a graphic tool that can help a lot with more CV focused applications. 

Regards,

Stav