Detecting covered eyes

Hi, here to know about an approach to a problem of detecting covered eyes, the eyes could be covered with anything, cloth, glasses etc. Basically what should be the approach in finding covered human body/face parts?

You could treat it as an image binary classification task.
Once you get the data for positive and negative classes, you could start with fine tuning a model as explained here.

Or would you like to “detect” the eyes with bounding boxes?
In that case, you would need to annotate your dataset with these boxes and could fine tune a detection model as explained here.