Device that updates recognition model and method of updating recognition model

1. An image processing device comprising:

one or more processors,

wherein the one or more processors function as:

an estimation unit configured to cause a category classification model learned based on learning data to estimate a category of input image data input to the category classification model and a score representing confidence of the category;

a display control unit configured to cause a display unit to display an image representing the input image data and the estimated category with the score below a threshold value estimated by the category classification model of the input image data;

an acceptance unit configured to accept a correction of the displayed category from a user; and

an updating unit configured to update the category classification model to an updated category classification model by adding image data of a corrected category based on the correction accepted by the acceptance unit and the corrected category to the learning data and performing additional learning, wherein the updated category classification model outputs the corrected category when the image data of the corrected category is input.

2. The image processing device according to claim 1 , wherein the estimation unit causes the estimation of the category, based on a threshold value to the score.

3. The image processing device according to claim 1 , wherein the score represents confidence of a first category, and the estimation unit causes the estimation of the category of the input image data having the score that is equal to or greater than a first threshold value, as the first category, and the category of the input image data having the score less than a second threshold value, as a second category.

4. The image processing device according to claim 1 , wherein the estimation unit causes the estimation of the category of the input image data having the score that is equal to or greater than a first threshold value, as a first category, and the category of the input image data having the score less than a second threshold value, as a second category.

5. The image processing device according to claim 4 , wherein the estimation unit estimates that the input image data having the score that is equal to or greater than the second threshold value and is less than the first threshold value, as a boundary data having no set category.

6. The image processing device according to claim 5 , wherein the display control unit causes the display unit to display the input image data estimated as the boundary data and the estimated category of the input image data caused by the estimation unit.

7. The image processing device according to claim 2 , wherein the display control unit causes the display unit to display a plurality of pieces of the input image data having the score near the threshold value and the estimated category of each piece of the input image data.

8. The image processing device according to claim 1 , wherein the display control unit causes the display unit to further display a distribution regarding the score of a plurality of pieces of the input image data.

9. The image processing device according to claim 1 , wherein the display control unit causes the display unit to further display a segment of the category regarding the score of a plurality of pieces of the input image data.

10. The image processing device according to claim 9 , further comprising:

an adjustment unit configured to adjust a threshold value to the score with a change of the segment of the category displayed by the display unit.

11. The image processing device according to claim 1 , further comprising:

a discrimination unit configured to discriminate whether the updating unit updates the category classification model.

12. The image processing device according to claim 11 , wherein the discrimination unit discriminates whether the category classification model is to be updated, based on a performance of the category classification model.

13. The image processing device according to claim 11 , wherein the discrimination unit discriminates whether the updating unit updates the category classification model, based on a level of wrong category data from the correction of the category.

14. The image processing device according to claim 13 , wherein the discrimination unit discriminates that the updating unit updates the category classification model, in a case where the level of wrong category data is equal to or greater than a threshold value.

15. The image processing device according to claim 13 , wherein the updating unit removes the wrong category data from target learning data and updates the category classification model.

16. The image processing device according to claim 5 , wherein the display control unit causes the display unit to display, in sequence, continuous boundary data having a difference value larger in the score than a predetermined value in order of the score in the boundary data.

17. The image processing device according to claim 1 , wherein the estimation unit causes the estimation of whether the input image data belongs to a normal category or an abnormal category.

18. The image processing device according to claim 17 , wherein the display control unit identifies and causes the display unit to display an abnormal part in the image representing the input image data estimated as the abnormal category.

19. An information processing method executed by an information processing device comprising:

estimating, using a category classification model learned based on learning data, a category of input image data input into the category classification model and a score representing confidence of the category;

displaying an image representing the input image data and the estimated category with the score below a threshold value estimated by the category classification model of the input image data;

accepting a correction of the displayed category from a user; and

updating the category classification model to an updated category classification model by adding image data of a corrected category based on the correction accepted by the accepting step and the corrected category to the learning data and performing additional learning, wherein the updated category classification model outputs the corrected category when the image data of the corrected category is input.

20. A non-transitory computer-readable storage medium storing a program for causing a computer to perform an information processing method comprising:

estimating, using a category classification model learned based on learning data, a category of input image data input into the category classification model and a score representing confidence of the category;

displaying an image representing the input image data and the estimated category with the score below a threshold value estimated by the category classification model of the input image data;

accepting a correction of the displayed category from a user; and

updating the category classification model to an updated category classification model by adding image data of a corrected category based on the correction accepted by the accepting step and the corrected category to the learning data and performing additional learning, wherein the updated category classification model outputs the corrected category when the image data of the corrected category is input.

21. An image processing device comprising:

one or more processors,

wherein the one or more processors function as:

an estimation unit configured to cause a category classification model learned based on learning data to output an estimated input image data input to the category classification model and a score representing confidence of the category, and to estimate the category of the input image data having the score that is equal to or greater than a first threshold value, as a first category, and to estimate the category of the input image data having the score less than a second threshold value, as a second category, wherein input image data not estimated as the first category or second category is determined as boundary image data;

a display control unit configured to cause a display unit to display an image representing one or more boundary image data whose score is equal to or greater than the second threshold value and less than the first threshold value;

an acceptance unit configured to accept category input for the displayed boundary image data from a user; and

an updating unit configured to update the category classification model to an updated category classification model by adding the boundary image data and the corrected category accepted for the boundary image data by the acceptance unit to the learning data and performing additional learning, wherein the updated category classification model outputs the corrected category when the image data of the corrected category is input.