The PVT is a type of
vision transformer that utilizes a pyramid structure to make it an effective backbone for dense prediction tasks.