rfinkelberg / stance Goto Github PK

Currently we have random print statements describing the state of the program. These need to be replaced with an actual logger

Currently some files have type hints and some do not. All of them should have consistent, mypy compliant type annotations

The way we do benchmark zones needs to be changed. We can't have a bunch of basically unnamed functions floating around for each motion

Maybe also examples of inputs/outputs?

This may just be researching what techniques/filters we can use to make the neural net either faster or better detect the skeleton

The entirety of get_points_from_image came from that one tutorial. In the header of the file we need to say that, credit the author, and link to the original code.

This includes creating benchmark zones, similarity function, and scoring function.

Each motion instance has:

• benchmark zones and functions to create skeletons for them
• skeleton similarity function
• scoring function