Table 1.

Semantic segmentation accuracy for the Mapillary Vistas dataset [3].

Pole	Street Light	Billboard	Traffic Light	Car	Truck	Bicycle	Motorcycle	Bus	Sign Front	Sign Back	Road	Sidewalk	Curb Cut
47.5%	35.8%	43.4%	62.8%	90.3%	70.4%	55.9%	59.1%	75.1%	69.5%	38.7%	88.6%	68.8%	14.7%
Plain	Bike Lane	Curb	Fence	Wall	Building	Person	Rider	Sky	Vegetation	Terrain	Marking	Crosswalk	mIoU
17.4%	37.3%	55.5%	55.0%	46.7%	86.6%	69.9%	47.3%	98.2%	89.7%	63.7%	53.5%	62.3%	59.4%