Sometimes you come across something that does two things:
- It completely invalidates what you were super passionate about 5 minutes ago.
- It allows you to skip a big step forward towards the next goal.
I am generally okay with 1. because that just happens, and is an integral part of learning.
2. is harder. 2. means that you have to actively adopt the new thing and make it your own. The choice is: find little details in their solution that you would have done better and complain about it (i.e. not being okay with 1.); or just accept it, change up, and move on.
I know the latter is the adult solution, but hmm… Let’s take a few days off to get back to a higher perspective.
Overall, the next steps are… To have a multi-camera setup maintain a world model. From MediaPipe’s perspective, this means running their holistic example on everything I can lay my hands on, and then find a way to integrate all measurements into a coherent representation.
That undoubtedly involves one of the newer SLAM implementations, and diving into the literature again.