首页>
外国专利>
LEARNING METHOD AND LEARNING DEVICE FOR STRATEGIC TRANSFORMING RGB TRAINING IMAGE SETS INTO NON-RGB TRAINING IMAGE SETS, TO BE USED FOR LEARNING OBJECT DETECTION ON OBJECTS OF IMAGES IN NON-RGB FORMAT, BY USING CYCLE GAN, RESULTING IN SIGNIFICANTLY REDUCING COMPUTATIONAL LOAD AND REUSING DATA
LEARNING METHOD AND LEARNING DEVICE FOR STRATEGIC TRANSFORMING RGB TRAINING IMAGE SETS INTO NON-RGB TRAINING IMAGE SETS, TO BE USED FOR LEARNING OBJECT DETECTION ON OBJECTS OF IMAGES IN NON-RGB FORMAT, BY USING CYCLE GAN, RESULTING IN SIGNIFICANTLY REDUCING COMPUTATIONAL LOAD AND REUSING DATA
The present invention provides a learning method and a learning apparatus for converting an RGB training image set into a Non-RGB training image set using a cycle GAN so that it can be used for object detection learning for an object of an image having a Non-RGB format, and It relates to the used test method and test apparatus. More specifically, in the learning method for converting an RGB image tagged with at least one correct answer information to a non-RGB image tagged with at least one correct answer information using a cycle GAN (Cycle Generative Adversarial Network), (a ) when the learning device acquires at least one first image having an RGB format, causes the first transformer to convert the first image into at least one second image having a Non-RGB format, and the first disk By making the reminator check whether the second image is an image having a primary Non-RGB format or an image having a secondary Non-RGB format to generate a (1_1) result, The head Non-RGB format is a Non-RGB format that has not undergone conversion from the RGB format, and the secondary Non-RGB format is a Non-RGB format that has undergone conversion from the RGB format, converting the second image into at least one third image having the RGB format; (b) when the learning device acquires at least one fourth image having the Non-RGB format, the second transformer converts the fourth image into at least one fifth image having the RGB format and the second discriminator generates a (2_1) result by checking whether the fifth image is an image having a primary RGB format or an image having a secondary RGB format, wherein the primary RGB format is It is an RGB format that has not undergone conversion from the Non-RGB format, and the secondary RGB format is an RGB format that has undergone conversion from the Non-RGB format, and causes the first transformer to convert the fifth image to the Non-RGB format. - converting into at least one sixth image having an RGB format; and (c) the learning apparatus, the first image, the second image, the third image, the fourth image, the fifth image, the sixth image, the (1_1) result and the (2_1)th ) calculating one or more losses with reference to at least some of the results, and learning at least a part of parameters of the first transformer, the second transformer, the first discriminator, and the second discriminator; It relates to a learning method and a learning apparatus comprising the, and a test method and a test apparatus using the same.
展开▼