Sythetic datasets of road images

The detector of elongated boundaries in the image, such as road marking lines, rails, etc., is an important component of the visual system of a highly automated vehicle (HAV). It is used by HAV to solve self-localization problems, maintain the position inside the lane, warning about lane departure.  Collecting and labeling data for solving these problems is always a time-consuming task.
We present two types of synthetic dataset of road images. The first type is aerial imagery dataset. It is intended to be used for road boundaries detector that works with bird's eye view road images. The second type of dataset is drawn road markings line on a black background. It is designed to optimize the parameters of elongated boundaries detectors that also works with bird's eye view road images but also first step of which is background suppression. Please note that not all parameters can be tuned but only those which affect steps followed by background suppression.

There are 2 folders "aerial_imagery_dataset" and  "drawn_road_markings_dataset", each corresponds of its own type of dataset. Each folder contains images and corresponding them markup in json files.  Pair image and corresponded markup will be called sample. "aerial_imagery_dataset" consists of 5336 samples, "drawn_road_markings_dataset" consists of 660 samples.
Markup files contain information of road markings coordinates in the image.