Recent work dealing with the problem of bridging the gap between simulated and real world data, especially in relation to image translation using Generative Adversarial Networks (GAN’s), predominately focused on translating simulated images into images, ideally, indistinguishable from real ones. This Bachelor-thesis, however, focuses on the opposite direction. Meaning the goal of the here presented image translation techniques is to translate realworld depth images into their simulated analogs. This approach is presented as an alternative option to bridge the gap between real and simulated data with its own drawbacks and challenges. These challenges as well as the benefits of this particular mapping will be discussed in this work in addition to the concrete methods and systems realizing this transformation. A discussion is held considering the major scenario analyzed in thiswork: Mapping the data in a supervised manner utilizing different Models (including primarily Auto Encoders and GAN’s). Furthermore each sub-scenario investigates the problem with data sets of diverse degrees of difficulty, starting with simple benchmark-sets, constructed in order to gain deeper insights into the underlying problems, and getting as close to real world data samples as possible. The Results are evaluated and compared, highlighting the strengths and weaknesses of each approach followed up by a conclusion for the whole inspected methodology.