Action Annotation

[asaeed9]

I have video data of a facility, and I would like to annotate this data for certain actions. People work with machines in this facility. It's dangerous work and we want to make sure that people do not put their hands on certain parts of the machine and interact with it by holding it from only allowed places. We want to design an AI solution that detects and highlights how an employee holds the machinery. AI solution would predict 'Good Hold' v 'Bad Hold'.

How best should I label this data such that I could train deep learning models that can help me predict my outcomes? Would appreciate any ideas.

Apologies in advance if this is not the right forum for this sort of question. Please point me towards the correct platform.