How do people convert an architecture meant for classification to segmentation?

Like I am a newbie and this whole thing is confusing me. I was going through the documentation of the Convolution Vision Transformer

architecture classification computervision convolution documentation newbie people segmentation through transformer vision