Method for adding prediction results as training data using AI prediction model

1 . A method for adding prediction results as training data using an AI prediction model, comprising steps of:

a) arranging a first image having a first opposing attribute, the first image being from an actual captured image of an arbitrary object, on a two-dimensional plane so as to be face-to-face with respect to a specific point on the two-dimensional plane;

b) arranging a second image having a second opposing attribute, the second image being from the image of the arbitrary object, on the two-dimensional plane so as to be orthogonal, with respect to the specific point, to the first image having the first opposing attribute;

c) receiving, by a computer system, image data of the first and second attributes respectively arranged on the two-dimensional plane;

d) predicting, by at least one AI prediction model, a first result for the image data of the first attribute received by the computer system;

e) predicting, by the at least one AI prediction model, a second result for the image data of the second attribute received by the computer system;

f) converting the first and second results predicted by each of the at least one AI prediction model into X and Y coordinates, to cumulatively store same; and

g) selecting data to be included in training by analyzing the X and Y coordinates of the cumulatively stored first and second results, and adding the selected data as training data.

2 . The method of claim 1 , wherein in the steps a) and b), the specific point is set to 50% or 0.5 which is a center value, and the upper, lower, left, and right maximum values are set to 100% or 1.

3 . The method of claim 1 , wherein in the step f), the X and Y coordinate values have a value of 0 to 100.

4 . The method of claim 1 , wherein in the step f), the cumulatively stored first and second results are brought later when analyzing the first and second results for a same target, and then analyzed again with regard to multiple occurrences of the same target.

5 . The method of claim 1 , wherein the at least one AI prediction model includes first and second AI prediction models, and in the steps d), e) and f), with regard to the image data of the first and second attributes, the first and second results are predicted by the first and second AI prediction models, respectively.

6 . The method of claim 1 , wherein in analyzing the cumulatively stored prediction results in the step g), the first and second results are analyzed by applying local outlier factor (LOF) analysis.

7 . The method of claim 1 , wherein in selecting the data in the step g), the data is selected by analyzing distribution of the X and Y coordinates of the first and second results using unsupervised learning and then classifying the first and second results into data results with relatively high cohesion and data results with low cohesion.

8 . The method of claim 1 , wherein in selecting the data in the step g), the data is classified into data closest to a center of each cluster, data farthest from the center, and data at an intermediate distance between the data closest to the center and the data farthest from the center, which are results of mean shift clustering.

9 . The method of claim 1 , wherein in selecting the data in the step g), the data is selected by setting conditions based on a distance from a center of each cluster, which are results of mean shift clustering.

10 . The method of claim 1 , wherein the at least one AI prediction model includes a single AI prediction model, and in the steps d), e) and f), with regard to the image data of the first and second attributes, the first and second results are predicted by the single AI prediction model.