Closed iamthemulti closed 1 year ago
This is not something that the current models have been trained to do. (As a side note, the models have been trained in various camera positions.)
Considering the case of two people in the frame, one way to do this could be using the pose information. E.g., if it's POV, then we will likely only see one full human skeleton in the frame. If it's not, then we are likely to see two human skeletons in the frame.
For example, to determine if a scene would be classified as POV or not.