Latent Space to Support Virtual 3D Models

This paper examines the potential of latent space in exploring digital realities. Latent space is a topological vector space of embedded data features so that, feature-wise, data that resemble one another are positioned more closely in latent space. That is, the distance between two points in latent space is a measure of similarity of their corresponding embedding data. Latent space has lower dimensionality than the original feature space from which the data points are drawn. We explore the relationship between feature space and the respective geo-physical qualities by statistically defining characteristics of the latent space. Through a case study, we demonstrate of how latent space can intervene in digital space using architectural, urban, or cultural properties from a geo-physical world. We conclude with a discussion of the potential application of latent space to the reconstruction of virtual 3D models.