Hi,

Lately I have been trying to control a robot with a puppet (kinda like a voodoo doll) and want to estimate the keypoints on the puppet to recognize the orientation of the head and the wheels so that I can transfer that as commands to the robot. I couldn’t find anything online about estimating keypoints on custom objects.

So that made me think that if a model could exist that can estimate the structural skeleton of objects, what could be it’s potential applications?

One application I can think of is recognizing the orientation of objects and not just detecting them in an image with a bounding box. What more could it be used for?

What are your opinions/thoughts on this?

p.s. I am thinking of this as my thesis for a research paper but just wanted to make sure it is something worth spending time for and knowing its potential applications.