Model Optimizer: convert custom model

[Nemes__Adam]

I want to use the following TensorFlow model in OpenVINO:

https://github.com/yeephycho/tensorflow-face-detection

Its a version of ssd_mobilenet, but not one of the supported topoliges.

I tried different --transformations_config options but i got one of the following errors:

[ ERROR ]  Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.middle.PartialInfer.PartialInfer'>): Stopped shape/value propagation at "Postprocessor/BatchMultiClassNonMaxSuppression/map/while/Slice_1" node.

[ ERROR ]  Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.middle.PartialInfer.PartialInfer'>): Stopped shape/value propagation at "BoxPredictor_0/ClassPredictor/BiasAdd/Reshape" node.

[ ERROR ]  Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.middle.PartialInfer.PartialInfer'>): Stopped shape/value propagation at "BoxPredictor_0/Reshape_1" node.

I used "--tensorflow_use_custom_operations_config" 

However i have limited information about the original model so i'm not sure which JSON use as paramereter,

and I DONT HAVE THE pipline.config file of the model.

What is the best way to convert a custom model?

Is it a good way to edit the JSON files from "deployment_tools\model_optimizer\extensions\front\tf" to fit the custom model?

[SuryaPSC_Intel]

Hi Adam,

You can try generating json and config files for your custom model taking the standard json files provided with Openvino and config files from tensorflow as reference.

Best Regards,

Surya

[Nemes__Adam]

Hi Surya!

Ok, so I'm on the right track.

Please explain "generate" is there a tool for that, or i just should edit them in text editor?

Thanks

Adam

[SuryaPSC_Intel]

Hi Adam,

OpenVINO does not provide any tool to generate json and config files. You can create your own taking reference from existing files.

Best Regards,

Surya