Semantic Segmentation

pixels-to-pixels

The output size is same as the input size.

end-to-end

One step to finish the whole process. The approach does not make use of pre- and post-processing com- plications, including superpixels, proposals, or post-hoc refinement by random fields or local classifiers.