DIGITS 3.0: Is scenelabeling supported?

Hi,

is it possible to train deep neural networks for scenelabeling with DIGITS?
Instead of one label per image, we need one label per pixel (scenelabeling).

Kind regards!