How to analyze classroom behavior using computer vision and pose estimation?

I am trying to build a computer vision system to analyze classroom behavior from surveillance cameras.

The goal is to automatically detect several behavioral indicators such as:

- student attention

- classroom participation

- social interaction between students

- activity level

My current idea is to use an object detection model such as YOLOv8 to detect students and then apply pose estimation (for example MediaPipe Pose or OpenPose) to analyze body movements and interactions.

However, I am not sure about the best pipeline for this type of task.

Questions:

1. What is the common approach for analyzing human behavior in a classroom using computer vision?

2. Should pose estimation be used after object detection, or is there a better architecture?

3. Are there any open-source projects or research papers that implement similar systems?

Any suggestions or references would be appreciated.