System and method for detection of road features based on artificial shadow data

A method for detecting one or more road features, comprising:

• receiving a first image associated with the one or more road features;
• detecting the one or more road features based on providing the received first image to a machine learning (ML) model trained at least in part on image data and artificial shadow data, wherein random shapes of shadows are determined based on a Bezier curve to generate the artificial shadow data;
• generating an output for the detected one or more road features included in the received first image;
• and
• updating a map database based on the generated output.

A non-transitory computer-readable storage medium carrying one or more sequences of one or more instructions which, when executed by one or more processors, cause an apparatus to perform operations comprising:

• receiving image data associated with one or more road features;
• determining a set of shadowed regions from one or more images of the received image data;
• segmenting the set of shadowed regions in the one or more images of the received image data from non-shadowed regions in the one or more images;
• processing the set of shadowed regions to obtain a set of shadow characteristics associated with the set of shadowed regions;
• processing the set of shadowed regions to determine texture information of the set of shadowed regions;
• determining random shapes of shadows based on a Bezier curve;
• generating artificial shadow data based on combination of the obtained set of shadow characteristics of the set of shadowed regions, the determined texture information of the set of shadowed regions, and the determined random shapes of shadows;
• and
• augmenting the image data to train a machine learning (ML) model based on the generated artificial shadow data.