Method and system for detecting changes in road-layout information

A method for training an artificial neural network for detecting changes in road information, the method performed by at least one processor of a computer system, the method comprising:

• training an adversarial learning artificial neural network to perform domain adaptation between a captured road image and a projected road-layout map by converting the captured road image and the projected road-layout map into first intermediate data and second intermediate data of a same feature space, respectively;
• training a metric learning artificial neural network to extract features of the captured road image and the projected road-layout map based on the first intermediate data and the second intermediate data, and calculate a distance between the extracted features as a similarity between the captured road image and the projected road-layout map;
• and
• training a change detection artificial neural network to detect presence or absence of changes in road information on the projected road-layout map based on the calculated similarity.

A system for detecting changes in road information, comprising:

• a domain adapter configured to convert a captured road image and a projected road-layout map into first intermediate data and second intermediate data of a same feature space, respectively, to perform domain adaptation between the captured road image and the projected road-layout map;
• a similarity determiner configured to calculate a similarity between the captured road image and the projected road-layout map based on the first intermediate data and the second intermediate data;
• and
• a change detector configured to detect presence or absence of changes in road information on the projected road-layout map based on the calculated similarity.