I am trying to run model given in this link on openvino.
As per I understand the liveliness detector in the link uses a custom trained keras model, I have used openvino sdk but didn't find any code which could be used for inference of custom model.
Any help on step to getting it working on openvino would be really helpful.
To convert your custom model. Kindly refer to convert tensorflow models.
Since this model is a Keras model, you will have to convert it to a frozen tensorflow model before feeding it to model optimizer.
I got the model converted to tensorflow model and used the mo_tf.py file to create openvino model.
Which sample code should I use to get the model working? In case no sample code are available are there any guides on building the sources for custom model?
I have worked with some of the sample application shared with sdk, I notice that sample codes has only code for converting the image to input to blob and pushing it to a list of blobs, and submitting the list of blobs for inference by calling function submitRequest.
The code for inference engine/loading the model is abstracted (Which I assume needs to be modified to fit in the newly created model by us). Not sure if the sources are open and provided with SDK..
Can you please let me know if my understand is correct?