Learning method and learning device for generating a virtual feature map having the same or similar characteristics as a real feature map by using GAN applicable to domain adaptation used in virtual driving environment, and test method using the same And test equipment

摘要

PROBLEM TO BE SOLVED: To alleviate the problem caused by training image set of non-RGB format by converting the training image set of RGB format to non-RGB format through cycle GAN which can be applied to domain adaptation. A learning method for deriving a virtual feature map from a virtual image having the same or similar characteristics as a real feature map derived from a real image using a GAN including a generation network and a discrimination network, the learning device. Corresponds to the output feature map by applying a convolution operation to the input image with the generation network to generate an output feature map having the same or similar characteristics as the real feature map, and the first loss unit. And generating a loss with reference to the evaluation score generated by the discrimination network. [Selection diagram] Figure 2