증강현실에서의 현실감 있는 가상 모델 가시화를 위한 Texture의 생성에 관한 비교 연구

초록

The virtual model reflects the reality and gives the user a sense of understanding and immersion in the model. For example, in a shipyard, a virtual model of a digital twin can increase work productivity by giving workers an immediate view of the model of the scene. However, Augmented reality virtual models are insufficient to describe reality of real model and there is a sense of heterogeneity, which reduces the objective commitment of work model and the reality of work environment. Although many methods to give a sense of reality to texture have been studied in the past, they have been confirmed that the method using Generative Adversarial Networks (GAN) is more efficient than other methods. GAN is a field of learning non-intelligence that can learn the image relation between two domains and transform it into a desired form. In this study, we have selected the most appropriate technique based on GAN to map the learning results to the virtual model. Based on these results, we expect to see various cases of shipbuilding marine industry using GAN.

키워드