Difference between init_inference and get_pose methods

[JMRMEDEV]

Doubt

What's the diference between init_inference and get_pose?, according to the comments on code, I see that both return an estimated pose.

[sneakers-the-rat]

@gkane26 can answer this better than I can, but quick answer since it's in my notifications -

init_inference prepares the model for use, creating the tensorflow objects, etc. depending on the type of model that's used. Some model types (tflite) need a frame in order to prepare the model, but it looks like it's just used to get a shape so it looks like that could be relaxed to also accept a shape tuple. Think the docs could spell that out a little clearer

if a frame is passed to init_inference, it calls get_pose, in which case it would return an estimated pose, otherwise it returns None.

get_pose is the main method for going frame -> pose.

it looks like there should be a check for initialization in get_pose so it raises a descriptive error if called before init, or else a little more permissive might be to auto-init with a warning saying "if u don't want to have first-frame lag call init first" or something. i'd be happy to PR that & docs if that sounds good.

[AlexEMG]

Yes, please PR into the docs, @sneakers-the-rat.

[sneakers-the-rat]

closing this to make a consolidated docs requests tracking issue