Robustness measurement device, robustness measurement method, and storage medium
AssigneesKabushiki Kaisha Toshiba, Toshiba Digital Solutions Corporation
InventorsTomoya TSURUYAMA, Hidetaka OHIRA
Life of this patent
Estimating the term…
Abstract
A robustness measurement device of an embodiment includes an acquirer, a noise adder, an inference result group generator, and a robustness measurer. The acquirer acquires an input image. The noise adder adds noise to the input image to generate at least one noise-added image by adding noise to the input image. The inference result group generator generates an inference result group by grouping mutually associated inference results among inference results for at least one of the input image or the noise-added image. The robustness measurer measures robustness of a test target model on the basis of the inference result group.
Claims12
A robustness measurement device comprising:
- an acquirer configured to acquire an input image;
- a noise adder configured to add noise to the input image to generate at least one noise-added image;
- an inference result group generator configured to generate an inference result group by grouping mutually associated inference results among inference results for at least one of the input image or the noise-added image;
- and
- a robustness measurer configured to measure robustness of a test target model on the basis of the inference result group.
A robustness measurement method comprising:
- acquiring, by a computer, an input image;
- adding, by the computer, noise to the input image to generate at least one noise-added image;
- generating, by the computer, an inference result group by grouping mutually associated inference results among inference results for at least one of the input image or the noise-added image;
- and
- measuring, by the computer, robustness of a test target model on the basis of the inference result group.
A computer-readable non-transitory storage medium storing a program for causing a computer to:
- acquire an input image;
- add noise to the input image to generate at least one noise-added image;
- generate an inference result group by grouping mutually associated inference results among inference results for at least one of the input image or the noise-added image;
- and
- measure robustness of a test target model on the basis of the inference result group.
Dependencies are inferred from the claim text.