Questions about data labelling- Stage II

[JasonSun-HUST]

Hey, thanks for this amazing work that means a lot to volleyball lovers.
I'm wondering if the action recognition could be more specific? For example, could the service category be further divided into jump serving and floating, or could the spike category go into tipping or other types of third contact? If not, is this the limitation of the CV models or just the limitation of data labelling?