首页>
外国专利>
GAN LEARNING METHOD AND LEARNING DEVICE FOR RUNTIME INPUT TRANSFORMATION OF REAL IMAGE ON REAL WORLD INTO VIRTUAL IMAGE ON VIRTUAL WORLD TO BE USED FOR OBJECT DETECTION ON REAL IMAGES BY USING CYCLE GAN CAPABLE OF BEING APPLIED TO DOMAIN ADAPTATION
GAN LEARNING METHOD AND LEARNING DEVICE FOR RUNTIME INPUT TRANSFORMATION OF REAL IMAGE ON REAL WORLD INTO VIRTUAL IMAGE ON VIRTUAL WORLD TO BE USED FOR OBJECT DETECTION ON REAL IMAGES BY USING CYCLE GAN CAPABLE OF BEING APPLIED TO DOMAIN ADAPTATION
The present invention relates to a method for learning a runtime input transformation that converts a real image into a virtual image using a cycle GAN that can be applied to domain adaptation, which can be performed in a virtual driving environment, and can be performed in (a) (i) 1 Let the transformer transform the first image into the second image, (ii) (ii-1) cause the first discriminator to generate the 1_1 result, and (ii-2) Causing the second converter to convert the second image into a third image having characteristics identical or similar to the actual image; (b) (i) causing the second converter to convert the fourth image into a fifth image, (ii) (ii-1) causing the second discriminator to produce a second_1 result, and (ii-2) Causing the first converter to convert the fifth image into a sixth image; And (c) calculating the loss. This method can reduce the difference between the virtual and the reality and the annotation cost.
展开▼
