Thanks. I think in a few years we'll see devices that can do everything locally, and cost a reasonable amount. For now, like others have said, some of the models for pose are too complex.