Data augmentation system, data augmentation method, and information storage medium

1. A data augmentation system, comprising:

at least one processor,

at least one memory, the memory having instructions stored thereon which causes the at least one processor to:

input, to a machine learning model configured to perform recognition, input data;

identify a feature portion of the input data to serve as a basis for recognition by the machine learning model in which the input data is used as input;

acquire processed data by processing at least a part of the feature portion; and

perform data augmentation based on the processed data.

2. The data augmentation system according to claim 1 , wherein the memory having instructions stored thereon further causes the at least one processor to:

select a part of the feature portion as a portion to be processed; and

acquire the processed data by processing the selected portion to be processed.

3. The data augmentation system according to claim 2 , wherein the memory having instructions stored thereon further causes the at least one processor to select the portion to be processed based on a score calculated for each area in the feature portion.

4. The data augmentation system according to claim 2 , wherein the memory having instructions stored thereon further causes the at least one processor to:

select a plurality of portions to be processed that are different from each other;

acquire a plurality of pieces of processed data based on the selected plurality of portions to be processed; and

perform the data augmentation based on the plurality of pieces of processed data.

5. The data augmentation system according to claim 4 , wherein the memory having instructions stored thereon further causes the at least one processor to randomly select from the feature portion the plurality of portions to be processed.

6. The data augmentation system according to claim 1 , wherein the memory having instructions stored thereon further causes the at least one processor to:

identify a plurality of feature portions;

acquire a plurality of pieces of processed data based on the plurality of feature portions; and

perform the data augmentation based on the plurality of pieces of processed data.

7. The data augmentation system according to claim 1 ,

wherein the input data is an input image to be input to the machine learning model, and

wherein the memory having instructions stored thereon further causes the at least one processor to:

identify a feature portion of the input image;

acquire a processed image by processing at least a part of the feature portion; and

perform the data augmentation based on the processed image.

8. The data augmentation system according to claim 7 , wherein the memory having instructions stored thereon further causes the at least one processor to:

acquire the processed image by performing mask processing on at least a part of the feature portion; and

perform the data augmentation based on the processed image on which the mask processing has been performed.

9. The data augmentation system according to claim 7 , wherein the memory having instructions stored thereon further causes the at least one processor to:

acquire the processed image by performing inpainting processing on at least a part of the feature portion; and

perform the data augmentation based on the processed image on which the inpainting processing has been performed.

10. The data augmentation system according to claim 1 , wherein the memory having instructions stored thereon further causes the at least one processor to identify the feature portion based on a result of the recognition output from the machine learning model.

11. The data augmentation system according to claim 10 ,

wherein the machine learning model is a model including at least one or more convolutional layers, and

wherein the memory having instructions stored thereon further causes the at least one processor to identify the feature portion further based on a feature map output from the at least one or more convolutional layers.

12. The data augmentation system according to claim 1 ,

wherein the machine learning model is configured to output a result of the recognition and an activation map for the result of the recognition, and

wherein the memory having instructions stored thereon further causes the at least one processor to identify the feature portion based on the activation map.

13. The data augmentation system according to claim 1 ,

wherein the machine learning model has a teacher data set including a plurality of pieces of teacher data learned therein,

wherein the input data is included in the teacher data set, and

wherein the memory having instructions stored thereon further causes the at least one processor to perform the data augmentation by adding teacher data including the processed data to the teacher data set.

14. The data augmentation system according to claim 1 , wherein the memory having instructions stored thereon further causes the at least one processor to generate a heat map from the input data using a Grad-CAM method, and wherein the heat map is used to identify the feature portion.

15. A data augmentation method, comprising:

inputting, to a machine learning model configured to perform recognition, input data;

identifying a feature portion of the input data to serve as a basis for recognition by the machine learning model in which the input data is used as input;

acquiring processed data by processing at least a part of the feature portion; and

performing data augmentation based on the processed data.

16. Anon-transitory information storage medium having stored thereon a program for causing a computer to:

input, to a machine learning model configured to perform recognition, input data;

identify a feature portion of the input data to serve as a basis for recognition by the machine learning model in which the input data is used as input;

acquire processed data by processing at least a part of the feature portion; and

perform data augmentation based on the processed data.