Pattern recognition based on information integration

摘要

A method for recognizing a primitive in an image includes recognizing at least one primitive in the image to obtain at least one candidate shape of the at least one primitive, which at least one candidate shape has a respective confidence; determining whether the recognizing of the at least one primitive has a potential error based on the confidence; obtaining auxiliary information about the at least one primitive from a user in response to determining that the recognizing has the potential error; and re-recognizing the at least one primitive at least in part based on the auxiliary information.

主权项

1. A method for recognizing a primitive in an image, the method comprising:
recognizing, with a processing device at least one primitive in the image to obtain at least one candidate shape of the at least one primitive, the at least one candidate shape having a respective confidence; determining, based on the confidence, whether the recognizing of the at least one primitive has a potential error, by determining if the respective confidences of at least two of the at least one candidate shape are higher than a first predetermined threshold, and a difference between the respective confidences of the at least two candidate shapes is lower than a second predetermined threshold; in response to determining that the recognizing has the potential error, displaying a user-interface and prompting the user to hand-draw the at least one primitive in said user-interface, wherein the hand-drawn primitive is used as auxiliary information; and re-recognizing the at least one primitive at least in part based on the auxiliary information.