METHOD FOR SELECTING LEARNED MODEL, METHOD FOR GENERATE TRAINING DATA, METHOD FOR GENERATE LEARNED MODEL, COMPUTER AND PROGRAM

摘要

To provide a method for selecting a learned model having high accuracy and the like according to an object, in object detection.SOLUTION: A learned model related to machine learning of an image is selected. Each image has a possibility of including three kinds or more of objects. A computer 10 acquires correct answer data A in which each object was detected, about a plurality of images I3 for a correct answer including a plurality of correct answer images I5 for learning and a plurality of correct answer images I6 for verification, generates learned models M1 - M6 according to a kind set for each kind set of the object, generates data B1 - B6 for verification for each kind set, and selects a model which is the most accurate in the data B1 - B6 for verification of the learned models M1 - M6 according to the kind set, for each kind of the object.SELECTED DRAWING: Figure 9