Come up with strategy for model deployment

[nashirj]

I am not sure how well the model will work if it is running on the raspberry pi (while a whole bunch of other things are going on). We can look into using a NVIDIA Jetson Nano, Google Coral, Intel NCS, or some equivalent just to host the model and give it some GPU capability and use TorchServe to serve our model and interact with this through the provided REST API

[ryPattillo]

Just realized , we could use one of the NCS that Stefan gave me for this.