ASR speech_sample Kaldi models and Open Zoo

[nikos1]

Hello,

We would like to evaluate the speech sample in OpenVino R3 using the existing Kaldi model referenced in the documentation ( https://software.intel.com/en-us/articles/OpenVINO-IE-Samples ).

There is a note that the Kaldi models will be available in the OpenVino model zoo

NOTE: wsj_dnn5b_smbr.nnet and other sample Kaldi models and data will be available in July 2018 in the OpenVINO Open Model Zoo.

What is the current status and availability on this? 

How can we access the OpenVino Model Zoo? 

Many thanks,

Nikos

[Severine_H_Intel · Accepted solution]

Dear Nikos, 

the Kaldi models are not available yet, I will keep you updated as soon as they are. 

Best, 

Severine

[Goncalves__Rodrigo]

Hello, is it available already?

[gaston-hillar]

Hi Severine,

I've checked and I still cannot find the models. The wsj_dnn5b_smbr.nnet file is not available in the model zoo yet.

Is it possible to have access to the models. Will it be possible to access the models before end of November 2018?

[gaston-hillar]

Today, I checked the documentation again and I found a new link to the models:

[gaston-hillar]

I forgot to paste the link in my previous comment in this thread for those that are searching for the models.

The included models are listed here: https://software.intel.com/en-us/articles/OpenVINO-InferEngine#automatic-speech-recognition

[gaston-hillar]

The following is the link to the models: https://download.01.org/openvinotoolkit/2018_R3/models_contrib/GNA/

[nikos1]

Great! Thank you gaston-hillar (Blackbelt)!

[gaston-hillar]

Nikos, it's great to know you read my comments. I was also waiting for these models to be available, and therefore, I thought it was a good idea to update this thread as soon as I found them. ;)

[Chen__Alex_G]

Hi , I'm trying to use mo.py to convert the wsj_dnn5b.nnet file which downloaded from that link  to OpenVINO IR files.  but failed.

can you help me to figure out which is wrong?

Model Optimizer arguments:
Common parameters:
    - Path to the Input Model:     /home/chengang/Downloads/wsj_dnn5b_smbr/wsj_dnn5b.nnet
    - Path for generated IR:     /home/chengang/openvino_models/ir/automatic_speech/FP32
    - IR output name:     wsj_dnn5b
    - Log level:     ERROR
    - Batch:     Not specified, inherited from the model
    - Input layers:     Not specified, inherited from the model
    - Output layers:     Not specified, inherited from the model
    - Input shapes:     Not specified, inherited from the model
    - Mean values:     Not specified
    - Scale values:     Not specified
    - Scale factor:     Not specified
    - Precision of IR:     FP32
    - Enable fusing:     True
    - Enable grouped convolutions fusing:     True
    - Move mean values to preprocess section:     False
    - Reverse input channels:     False
Kaldi specific parameters:
    - Path to the counts file:     /home/chengang/Downloads/wsj_dnn5b_smbr/wsj_dnn5b.counts
    - Removes the Softmax layer that is the output layer:     True
Model Optimizer version:     1.4.292.6ef7232d
[ ERROR ]  Model Optimizer is not able to read Kaldi model /home/chengang/Downloads/wsj_dnn5b_smbr/wsj_dnn5b.nnet. 
 For more information please refer to Model Optimizer FAQ (<INSTALL_DIR>/deployment_tools/documentation/docs/MO_FAQ.html), question #91. 
Error on or near line 83; exiting with status 

and the FAQ #91 is 

91. What does the message "Model Optimizer is not able to read Kaldi model .." mean?

There are multiple reasons why the Model Optimizer does not accept a Kaldi topology: file is not available or does not exist. Refer to FAQ #89.

[nikos1]

Could it be that wsj_dnn5b.nnet was corrupted during download?

What is the output of 

ls -l  /home/chengang/Downloads/wsj_dnn5b_smbr/wsj_dnn5b.nnet