Figure 15.

Results of semantic segmentation of the road. First row: some examples of road scenes in various situations. From left to right: City shadow mix scene with moving bicyclist and vehicle, back-road shadow mix with no traffic, urban city road clear sky with cars parked, tunnel dim with no traffic. Second row: the corresponding road ground truth annotations in yellow overlaid over the original image. Third row: road segmentation by our model FCNx. Row four: road segmentation by UC Berkeley FCN8.