Why model classifying in images into two classes. I want multi class classification.

I am trying this model on multiclass classification. My trained dataset contains four folders Happy, Sad, Neutral, Angry. The first problem it's reading every image from all folder when it is going to read angry or any other label image and second it's not taking the two labels only angry and sad it's taking not other two.
 Can anyone help me with that?
Eagerly waiting for the solution.